First off just a tag-based implementation of Ext4 overrides.

As we no more support JRuby, we can require therubyracer as a dependency.

This reverts commit aba71287.

Problem caused by the addition of --ext4-events option, which caused the
append_ext4_event_options method to no more behave as it did.

As I'm dropping JRuby support, there's no need for extra redirection
layer.

Not needed for day-to-day testing, a drag on performance.

IE7 & 8 will download this kind of URL twice, so better to use a script.

See also: http://stackoverflow.com/questions/4831741/can-i-change-all-my-http-links-to-just

Won't work in Ruby 1.8.

This way it'll play nicely when JSDuck is hosted over HTTPS.

Rewrite #! links into ?print= links.

The stylesheet assumes .x-panel-body div which wasn't present in print
page.

I use it rarely, so no need to waste time on creating that.

We're still defaulting to auto-detection, but this option can be used to
override the auto-detection.

Otherwise we need to pass around the ParallelWrap instance or config
for it, which is annoying.  Parallelization is more of an implementation
detail - passing it explicitly to every class that needs to make use
of it doesn't really fit to the interface of these classes.  We
shouldn't need to change the interface when switching from normal each
loop to parallel each.  So a singleton class it is.

Getting input from jsduck export, feeding output to Graphviz Dot program.

For example:

    ./bin/graph docs-export/ | dot -Tsvg > diagram.svg

To be in sync with the new header style used in docs.sencha.com.

Reimplement Guides#each_item so that the guide doesn't get passed
to the block when loading of it had failed.

Also make Logger#warn return false to be able to return with false
right after showing warning.

To allow easier changing of title of the produced HTML file.

The compare script now lists added, removed and changed members.

The change tracking is currently targeted towards configs. We
detect the following changes:  type, default value, visibility.

I refactored the script so it now finally has a sensible interface,
with options --ignore (for the optional ignored methods/classes file)
and --type (to limit output to just configs).

Use class and member names instead of indexes.

When documentation contains @since and/or @new tags, they always
take precedence over auto-generated ones.

Along with @since tags generate @new tags.

@new tag is added to any class/member that is available since the
very latest version.

Graphically the @new tags will be rendered as stars on yellow
background.

Consequently OldVersions module now becomes Importer.

Newer connect-mongo versions cause trouble. Needs investigation.