When using cookie-based auth, login / logout happens via HTTP requests
made by the "session" service and the auth service has to be explicitly
notified afterwards via `auth.clearCache()`.

When using OAuth-based auth on the other hand, login / logout happens
through the auth service itself, so there is no need to ask the auth
service to clear cached credentials afterwards.

Implement logging out in the OAuth authorization service when using a
first-party account.

This currently just causes any credentials cached in memory or local
storage to be forgotten and the user's profile to be reloaded. Once a
token revocation endpoint is available, the "Log out" button could also
trigger that as well.

Use correct request params when exchanging auth code for tokens

HTML5 Audio elements

Made changes suggested in code review:
https://github.com/hypothesis/client/pull/508#pullrequestreview-53812970
.

Tiny fix to the test to account for URLs that may have had uppercase
letters (the audio embed converts to lowercase to account for links to
.MP3 files rather than .mp3)

Fixed tests for audio elements.

Some API URLs are created by resolving relative URLs against the API
root. For this to work properly, the root URL has to include a trailing
slash.

The omission of this in the documentation caused confusion for eLife
devs.

The two most recent client releases failed to publish automatically
because the latest tags had not been fetched prior to running
`git describe` to check whether the git ref was a tag.

I'm not sure exactly what changed and why this used to work before,
but add an explicit `git fetch --tags` step.

eg. See https://jenkins.hypothes.is/job/client/job/master/483/

Remove the view switcher component

Since the h service now supports authorization codes rather than faking
them with JWT grant tokens, use the appropriate request parameters in
the call to exchange the auth code for tokens.

Remove the (feature-flagged) view switcher component.

We missed the chance to strike while the iron was hot on this one -
neither Dawa or I have time anymore to work on further design
iterations. I don't want to have dead code living behind a feature flag
so I'm removing it. The code will still be in git if we ever want to
revive it. Perhaps these tabs can be re-designed as part of a larger
re-design of the sidebar in the future.

I'll leave the GitHub issue about the usability issues with the existing
selection tabs open: https://github.com/hypothesis/product-backlog/issues/327

For the record the main remaining issues with this were:

1. An undesirable visual popping happens when the view switcher loads on
pages that have a lot of annotations. This also happens with the
existing selection tabs but is arguably more visible with the view
switcher.

2. When there are only two tabs (no orphans) the visual design of the
tabs doesn't make it as immediately obvious which is the currently
selected tab. The other tab may be mistaken for the selected one.

The TODO list was here:

https://github.com/hypothesis/product-backlog/issues/327#issuecomment-311073581

The pull requests for this view switcher were:

https://github.com/hypothesis/client/pull/429
https://github.com/hypothesis/client/pull/481
https://github.com/hypothesis/client/pull/482
https://github.com/hypothesis/product-backlog/issues/327

Adding ADR for ePub support

Adding onLayoutChange documentation

More cleaning up of the media-embedder tests

Cleaned up media-embedder tests

Added media-embedder tests for .mp3, .ogg, .wav files

Turns links to web-published .mp3, .ogg, and .wav files into HTML5
audio elements

Fix IE 11 regression in documents without a `<base>` element.

`Document#uri` and `Document#getMetadata` returned the base URI rather
than the document's real URL as the document location if the protocol of
the base URI and true document location were different. The reason for
this was to support the use case where the document is an iframe loaded
from a `blob:` URL, which Readium does for example.

This broke in IE 11 if the document did not have a `<base>` element
because the fallback `Node.baseURI` and `Document.documentURI`
properties used by `document-base-uri` do not exist in that browser.

This commit re-works the behavior of `_getDocumentHref` to be more
conservative and return the base URI only if a) the
document location does not have a whitelisted scheme and b) the base URI
is defined and does have a whitelisted scheme.

Additionally, the tests have been re-worked to test the public `uri`
method instead of the private `_getDocumentHref` method and use test
seams instead of proxyquire.

Code clean up (removed H5P embed test)

Added support for mp3 -> html5 audio and a primitive h5p embedder.

Use a URN derived from DC meta tags for creating equivalent doc links

“dc.relation.ispartof” is now used instead of “dc.source”
The goal of this is to identify a resource that’s part of another resource, therefore the more specific term fits better.

Use a URN derived from DublinCore meta tags for creating equivalent links across documents that provide ‘dc:source’ and ‘dc:identifier’ tags

This link is now the documentIdentifier for non-PDF cases for the document metadata. This metadata is shared with the sidebar, making this link appear in the list of links for the Search API calls