Various classes were not directly listed with the requires statement.

Additionally Sencha command complained about being unable to find
Ext.JSON, so excluding this, as it's within the ext core anyway.

It was moved before the other scripts to ensure that comments
enabling code is included before docs app code, but new that the
comments code is withing data.js, that's no more needed.

Now one can simply use these options instead of injecting special
comments-enabling JavaScript with --head-html.  The JavaScript side
is now also better as it accesses the commentsUrl and commentsDomain
properties of Docs.data object, not a lot more properties on the
root Docs object.

This way it's apparent to the user of the site why he gets served
with a particular version.

Instead of using the server-side PHP script, the browser detection
is done one the client side with a compact JavaScript snippet taken
from http://detectmobilebrowsers.com/

This allows the index page of the docs to be cached, lowering the
burden on servers.

Instead of crashing mysteriously, the circular @mixins and @extends
now result in fatal error message which clearly shows the
dependency chain that's causing the problem.

Move {head_html} before the app.js.

This way the comments settings will be available before
the rest of the app is loaded, especially the CommentsProxy which
assumes the Docs.comments* settings to be available.

Default Docs.Comments#getCount to return 0 when comments aren't enabled.

The monsterid was a bit too silly. Docs seem to require something
a bit more formal.

A cleanup of leftovers from old MongoDB comments database.

Extract TopList base class.

Define Tag and Target models using the new "comments" proxy which
helps us with prepending comments-specific stuff to URL.

Replace the duplicated code with a more generic solution.

Though it's now more complex, it'll allow us to add/remove/rename
cards in the future more easily.

This change breaks compatibility with comments API of previous
JSDuck release.  But the only effect of this is that the Users and
Targets lists on comments index page will be empty, so I consider
it OK to make this change as it will only make the comments index
page a bit less useful while the transition happens, but won't
break anything badly.

Rename Comments.getAllTags to .getTopTags, so it's in line with
.getTopUsers and .getTopTargets.

Ensure tags are sorted alphabetically in TagEditor (the backend
now returns them sorted by comment count).

For now the comment counts aren't shown in tags menu, but otherwise
it's working.

For now the list of tags is completely hardcoded.

Send tags along with comments from backend.

Render the tags in comments template.

Improved stylesheet for rendering all the items in top-right corner
of each comment.

For moderators, also render the add/delete buttons for the tags.

Implement the tags in nice relational fashion, but set up triggers
to avoid quering the tags table with each select of comments, and
instead precalculate a textual tags column inside the comments
table.

Also added a migration script.

Just rebuild the entire total-comments-per-class cache.
We could do it in a smarter way, but as it only needs to be
done when posting/deleting comments, it happens rarely enough.
And the performance is actually just good-enough; no noticable lag.

Use the common Comments.avatar method to generate the <img> tag,
giving us the same default avatar that we use everywhere.