Including the ones belonging to parent classes.
All methods are augmented with :member field that
says the class they belong to.

In top-level filter_classes each Class instance
also receives a hash of all classes - so that the
Class.parent() method can return instance of parent class.

Finally added some rudimentary tests for all this.

Introduction of ident/keyword separation broke this before,
now it's working again.

Another regex v/s division tokenization enhancement.

With that generalized the regex token detection - a division operator
can never follow a keyword, but regex literal can.

Both Page and Tree can now use these common methods without implementing
their own.

Giving warinings for other types.

The to_html method currently creates the top section of page.
Definitely quite buggy, but at least we have started it.

JsDuck.write_tree now also outputs empty Docs.icons object,
which keeps UI from crashing - but we should fill it with
actual icons-map in the future.

--output now specifies output directory.  The tree is now written to
  file instead of stdio.
--verbose turns on printing of parsing progress.

JsDuck::Tree creates from doc-object list a tree structure consisting
of Ruby Hashes and Arrays.  Another method in jsduck.rb converts this
to JSON and outputs.

For all this added a unit test, but with it's single test it's mostly
just a sanity check.

The previous debug output was removed to separate method.

jsduck script now reads filenames from command line, parses each one
of them and prints out tree of classes and their properties.

Not quite correct, but for now it's better for them to behave the same.

Again one of the most used @tags in ExtJS.

Only supporting it in class declaration, plus some explicit logic to
detect it when it appears after @constructor in class doc-comment.

Applies only to properties and methods.

This was one of the most used @tags in ExtJS not already supported by
JsDuck.

Class member arrays are now named the same as their tagnames.
So we access class configs with  class[:cfg]  rather than with
class[:cfgs]

All the complex comment and code analyzing is now done by Merger class,
the doc_comment_parser now just parses input into a linear array of at-tags.
The Merger completely replaces DocComment class and most of the complex
work in DocCommentParser.

Most of the tests are now also at top level tc_jsduck - so that we
mainly test the whole jsduck module, allowing the internal classes to be
more easily reorganized.

@ignore and @hide are now simply aliases of @private.

The results of DocComment and JavaScript parsing are merged together in
top-level parse method.

Devided files between lib/ and test/ directories.
Added Rakefile that runs all the tests.

The most important at-tags are now supported.