innerText is non-standard and not supported in Firefox < 45.

Use textContent instead.

Enable video embeds to be shown fullscreen

Improve a code comment

The 'allowfullscreen' attribute must be set on both
the embed's <iframe> and the parent sidebar <iframe>
when the embed's main script is loaded to enable
fullscreen support.

Add media embeds feature

Client-side, replace any links to YouTube or Vimeo videos in annotation
body texts with <iframe> video embeds for those same videos.

Behind a new feature flag.

markdown.coffee: after rendering the user's markdown text to HTML and
sanitizing it, compile the HTML string into a DOM element and insert it into
the DOM using innerHTML from JavaScript, instead of using Angular's
ng-bind-html to insert the HTML string into the DOM.

This means that our CoffeeScript code in markdown.coffee has the rendered
annotation body as a DOM element, and can use the DOM API to modify it before
inserting it into the DOM.

Take advantage of this to find video links in the HTML and replace them with
embeds. This is done by a standalone media-embedder.js module that
markdown.coffee calls.

This module is designed to allow functions for generating embeds for different
kinds of URLs to be plugged in. So far functions for a few different kinds of
YouTube and Vimeo URLs are implemented, but further functions for more
media types can be added.

jscs: Require `function () {}`

Change jscs's required style from `function()` to `function ()`.

We always require a space after the `function` keyword, and before {}
braces, but we do not allow a space between a function name and its ()
braces:

    var foo1 = function () {};  // Right.
    var foo2 = function() {};  // Wrong.
    var foo3 = function (){};  // Wrong.

    var foo4 = function foo4() {};  // Right.
    var foo5 = function foo5 () {};  // Wrong.
    var foo6 = function foo6(){};  // Wrong.

    function foo7() {}  // Right.
    function foo8 () {}  // Wrong.
    function foo9(){}  // Wrong.

This deviates from the Google style guide and jscs presets that we use.

Don't crash on non-JSON body

Enable showdown's literalMidWordUnderscores

This stops showdown from interpreting underscores in the middle of words
as <em> and <strong>, instead it treats them as literal underscores.

Example:

some text with__underscores__in middle

will be parsed as

<p>some text with__underscores__in middle</p>

Fixes #2598 (URLs with _'s in them were getting <em> tags inserted into
them, breaking the links).

Also see <https://github.com/showdownjs/showdown/issues/211>.

Small tweaks to the header for re-use in the blog

Don't update annotations with contents of others

When listening to the 'annotationUpdated' event, AnnotationController
needs to check that it's this annotation that the event refers to before
acting on it.

Fixes #2799.

Annotation controller domain model view model refactor

If the user creates annotations while signed out and then signs in, we
want those annotations to still be present in the sidebar after sign in,
rather than being lost.

Commit 7a1b5ca8d99252d3a086a8a8103fca4702521d06 had inadvertently broken
this.

The problem is that app.coffee destroys all AnnotationController
instances on sign in, and only creates new instances for those
annotations that have drafts.

7a1b5ca8d99252d3a086a8a8103fca4702521d06 removed a spurious call to
saveToDrafts() in vm.edit() but it turns out that this call was
(apparently inadvertently) the only thing that was causing drafts of
annotations created while signed out to be saved, and hence the only
thing causing these annotations to persist across sign ins.

So add a listener for the USER_CHANGED event and save drafts of
annotations on sign in, just before the AnnotationControllers get
destroyed.

Update AnnotationController's view model on the "annotationUpdated" event.

This fixes an issue that "live updates" were not being shown on annotation
cards.

The "annotationUpdated" event is fired by AnnotationController itself after
successfully saving updates to an already-saved annotation. It's also fired by
annotationMapper after received an updated annotation from the server (that was
updated by a different client - the "live updates" feature).

Subscribing to this event allows one function in AnnotationController to handle
updating the view model in both cases.

Update AnnotationController's view model (vm) after successfully saving
a new annotation to the server for the first time.

This fixes two issues:

1. vm.annotationURI (which depends on the annotation's id, which is not
present until the save to the server succeeds) was not being set
correctly after successful save, so the annotation permalinks were not
correct until you reload them.

2. If you created a highlight while logged out, then logged in,
(AnnotationController automatically saves the highlight to the server
for you on login) the "Only Me" icon was not being shown on the
highlight until you reloaded it because vm.isPrivate wasn't set.

We only need to save an annotation to drafts when the focused group
changes while editing it, and not at other times (e.g. when opening the
editor??).

We only need to restore an annotation from drafts when editing it, and
not at other random times.

Also, don't call updateViewModel() in init() after calling edit(),
because this will overwrite the draft changes that you just restored.

Don't change the value of domainModel.group when the focused group changes.

Instead, change it to groups.focusedGroup() in updatedDomainModel() before the
annotation is sent to the server.

This fits with the idea that domainModel is a read-only copy of the annotation
that we received from the server, and vm is the read-write copy. But there's no
vm.group as this would always be the same as groups.focusedGroup().

Recent changes broke the reverting of unsaved changes to annotations
(AnnotationController.vm.revert()) when the user presses the _Cancel_ button on
the annotation editor. The annotation's privacy was not being restored to its
previously saved value.

We can't set domainModel.permissions when the user changes the privacy of an
annotation that they're editing, because if the user clicks _Cancel_ we need
to revert the privacy back to its previous value (which we've just overwriten).

So instead add a new vm.isPrivate boolean and set that instead:

- vm.isPrivate() method (which returned true or false based on the value of
  domainModel.permissions) is replaced with a vm.isPrivate boolean.

- updateViewModel() now sets vm.isPrivate according to the value of
  domainModel.permissions.

- updatedDomainModel() nows sets domainModel.permissions according to the
  value of vm.isPrivate.

- setPrivacy() nows sets vm.isPrivate instead of domainModel.permissions.

- We no longer change domainModel.permissions when the focused group
  changes (it will continue to hold shared permissions for the previous group,
  but these will be updated by updatedDomainModel() before saving them to the
  server).

  vm.isPrivate doesn't need to be changed when the focused group changes
  because its value is independent of the focused group.

- saveToDrafts() now saves vm.isPrivate instead of vm.isPrivate()

- restoreFromDrafts() now restores vm.isPrivate instead of
  domainModel.permissions

These were being converted from view model to domain model format before
being stored in drafts, and back again after being retrieved, which of
course is pointless.

vm.document was a variable containing data duplicated from domainModel (not an
exact copy, but data extracted from domainModel anyway).

It's just easier to have a method for getting this - saves having to manage
another variable and update it at the correct times.

In making this change I noticed that there were several tests for
updateViewModel() that were actually tests for extractDocumentMetadata()
(which updateViewModel() was calling). These tests are no longer valud since
updateViewModel() no longer calls extractDocumentMetadata(), and there are
already duplicate unit tests of extractDocumentMetadata() itself.

This leaves the describe('updateViewModel()') empty of tests, and since
updateViewModel() is now pretty trivial, I decided just to delete the
describe() rather then leave it empty or think up some tests to add to it.

I'm not sure this description adds much, and the comment about using
annotationMapper for persistence is wrong.

Clarify what this object actually is: the read-write variables for the
templates to write via ng-model.

This variable is just duplication of vm.annotation.tags and is not used by the
templates.

A bug had crept in where if you changed the privacy of an annotation
that you were creating or editing (by changing the selection in the
_Post_ dropdown) then the privacy would not change and your text and tags edits
would be deleted.

This was happening because when you change the selection in the Post
dropdown then onSetPrivacy() makes a change to the domain model which
triggers onDomainModelChange() which copies the domainModel over
vm.annotation, overwriting your changes.

To fix this get rid of onDomainModelChange(). These sorts of functions
(ones that are triggered by events or changes) create confusion by
making it difficult to predict or trace the flow of execution, and I
can't see why onDomainModelChange() needs to exist.

This does deliberately remove one feature:

If you had saved edits for an annotation in the drafts service (for
example if you edited an annotation, made some changes, then changed the
focus to a different group without saving the annotation) and _then_
someone using a different instance of the client (e.g. yourself on a
different device, I guess) saved a change to that same annotation to the
server, then this change would be communicated from the server to the
first client (via the WebSocket I guess) and the first client would
*delete your saved edits from the drafts service*.

This behaviour was deliberate but I'm not sure why it was considered a
good idea, and removing it simplifies things.

Deduplicate and correct code in AnnotationController for deciding
whether an annotation is a reply.

Remove all the properties that don't need to be in vm.annotation from
vm.annotation.

This means removing all the read-only properties (id, target, updated, user)
and leaving only the read-write ones that the templates write via `ng-model`
(vm.annotation.tags and vm.annotation.text).

The templates still do need read-only access to id, target, updated and user,
but rather than _duplicating_ these values by copying them from domainModel
to vm.annotation, instead add accessor methods vm.id() etc that just return
the values from domainModel.

This has two advantages:

1. Data is not duplicated between domainModel and vm.annotation.
2. It's clear that these fields are read-only, since the templates literally
   can't write them, at least in the case of id, user and updated

   target is unfortunately a mutable object so the templates could still
   mutate it. vm.target() could return a copy to avoid this but I haven't gone
   that far.