Migrate all users to using the InputModel.

Instead of WindowModel defining lots of methods on the off-chance
that the window is writable, pull them all into a separate
optional InputModel.

For ease of transition, FrameContainer will just implement both,
but eventually the input handling should be pulled out into
its own class.

These will be going away shortly.

Despite its name and description, the WindowManager relied on
each FrameContainer having a reference to its parents and all
its children.

The WindowManager now stores this information itself, and so
the FrameContainer methods can be gradually phased out.

Closes #426

Issue #426

Formats defined in YAML can specify 'timestamps: false' to hide
the default timestamp for lines associated with that event.

Closes #633

Also make event bus errors have a better message.

Closes #627

Closes #622

This enables release tracking in sentry.

Issue #622

This will allow plugins to supply their own format providers
while still allowing the user to override them using the
main format.yml file.

Issue #622

This release contains some performance improvements compared
to the version we were using.

Error handlers are now passed in at construction time, so
there was a small bit of refactoring required.

With numeric formatting gone, all output is either added direct
to a window (which will be changed soon) or done through
displayable events. Message sinks are now completely unused,
as they don't affect displayable events.

Issue #426

This allows events to iterate over a collection property, and
adds formatting for whois results. It needs a lot of tidying,
but it's a start...

This allows formats that may be multi-line (whois results,
probably MOTD results at some point, etc...) to show a header
and footer around the actual content.

This allows us to remove a fair chunk of the notification
handling functionality in FrameContainer.

Issue 426

I think these will eventually go away entirely, once everything
is event-bus'd, but I'm not entirely sure how that works any
more :(

Split out error description getting into its own method. Other
minor changes.

We have a path, it doesn't really make sense to take the filename,
then take the directory, then create a new path using those...

The logic to get a correctly typed setting given a class is
pretty general purpose, and will be useful elsewhere. Pull
it out of ConfigBinder and add some tests for it.

Dagger 2.0 appears to require each method in a module to have a
unique name. In some cases we've used overloads with the same name.

Also, remove a bunch of boilerplate javadoc.

Issue #601

We were raising notices as all coming from the local user, not
the person that sent the notice. This is... unideal.