This commit adds a streamer service that abstracts interaction with the
streamer websocket.

It's not clear to me why we were using this rather complicated method of
generating a UUID. Suffice to say that simply calling uuid.v4() has much
the same effect.

This isn't used anywhere else, so simplify the interface of Socket by
inlining a call to get the client ID.

... to cfac672e (typed-packaging branch).

Fixes #1788.

... when the bucket bar is being updated
before the document access strategy becomes
available. (This sometimes happens when using
PDF.js in the Chrome extension.)

Raise a error if the ICU ElasticSearch plugin is not installed

Remove an unused event

The original commit that introduced this (by @csillag) does not at
all describe why it was added. Nothing in this project listens
for it.

O_O

Postpone the scanning

... to the point when the panel is already ready.

Hopefully, this makes the sidebar connect and appear sooner.

We can't delay in any further, because setting the
visible highlights (which happens in the next line)
will trigger a bucket bar (a.k.a. heatmap) refresh,
and the heatmap uses information produced by the scanning.

Update PDF plugin

... to 18d8f72a (typed-packaging branch),
which brings support for the PDF.js version
shiping with FF35.

Move thread ng-show conditionals onto parent elements

Fix double loading on install

Rather than the thread-reply and thread-load-more elements having
one set of conditions on the parent and one on the child we now ensure
that the entire element is always shown/hidden.

Remove dynamic bucket from BucketBar

Rename Heatmap -> BucketBar.

Ref #1236

Don't use target="_blank" for feedback email.

Consolidate code paths around deciding about PDF.js

Rename rpc methods to be more consistent

Annotator's PDF.js plugin replaces the document plugin
for PDF.js, so when we are using the PDF plugin, we
don't want to load the document plugin.

For this, we are removing it from the options, in
presence of PDF.

This commit changes how we test for PDF.

Earlier, we were simply testing for the presence of the
global PDF.js object, but it's better to actually
use the test already implemented inside the PDF plugin.

This has two benefits:
 * If the PDF plugin is not present, we won't
   mistakenly  remove the document plugin.
 * If the PDF plugin decides that it can't deploy,
   for whatever reason, we won't
   mistakenly remove the document plugin.

Between guests <-> sidebar
- `showViewer()` to  `showAnnotations()`
- `updateViewer()` to `updateAnnotations()`
- `toggleViewerSelection()` to `toggleAnnotationSelection()`

All these renames were to enforce encapsulation and these calls
are more like APIs. To guest side really do not have to know anything
what the sidebar has or does.

Renamed a few methods to be more consistent

(As previously discussed with @tilgovi)

Sidebar -> Document RPC methods:
 * Renamed `scrollTo()` to `scrollToAnnotation()` (to be consistent with `focusAnnotations()`

Guest methods:
 * Renamed `showViewer()` to `showAnnotations()` (to be consistend with `focusAnnotations()`)
 * Renamed `updateViewer()` to `updateAnnotations()` (same reason)
 * Renamed `toggleVIewerSelection()` to `toggleAnnotationSelection()` (same reason)

This reverts commit eb3c486991d921a1f2342a289cb9755d9d76459e.

I don't think this fixed anything because the permissions plugin
should still have been setting the permissions to empty.

It is fine if the permissions are initially undefined.
Fix #1747

Create reindex.py, add warning on mapping conflict

Update License and Notice files

Highlight state considation