Remove an old restriction left over from the days when JSDuck didn't
show private members at all.

Only when the enum is an auto-expanded list of alias values
or when it's defined as array - only then show the warning.

The :config_file task generates initial simple sdk-vars.rb config.

The :configure runs both :get_extjs and :config_file, doing the
initial setup for running jsduck.

This should help tools like readability.com extract the contents
of the page.

So we detect the constructor the same way.

It's quite obvious what constructors do.

Also added a method Class.constructor? to check if a member is
a constructor function.

This ensures that when running the export repeatedly the order
of mixins/requires/etc will stay the same.  Especially in the
case of mixins which are often defined in an object literal which
gets turned into a Ruby hash that messes up the order quite a bit.

By doing the sorting in Export::Full the Renderer no more needs
to sort these things.

Make it work with renamed Json class.

Print also those classes that are exactly the same.

Only turn properties into configs when dealing with autodetected members.

The validation relied on user_id field being available in comment
data, but this field was in fact missing, resulting in the comment author
to never equal current user.

Moved bunch of code away from the humongous Docs.view.Comments.

The emailHash of user wasn't sent when user logged in.

Add note about replying to other people.

Attempt making the syntax help more visible.

Add a new version with a different name alongside it.

This way we can keep two versions of the API working at the same time,
which is needed when first deploying the backend code and then waiting
for the frontend code to get moved into live.

Added field to the top of both users and topics which allows filtering
the lists below them.

Every time the filtering happens, the previously existing selection is
cleared.  This is maybe not the best behavior, but that's the compromise
I arrived at to prevent several edge-case bugs.  The problem is, that
when you filter out a currently selected record, it becomes deselected,
but the deselect event doesn't fire to let us know...