Copying of touch build was done inside copy_touch2_images, which
wasn't really the correct place.  Also, all the copy_*_images
methods were removed, so I created a separate copy_touch_build
method.

Now all images can be marked up the same way - no need for special
syntax in guides.

Using this, JSDuck will automatically copy over images referenced
by {@img} tag.  No more need for extra copy task after running JSDuck.

The categories JSON file must now just contain an array, just like
the JSON files for videos, guides and examples do.

Simplicity FTW!

Filename always at the beginning, followed by line number and message.

JS files not needed if --welcome, --guides, --videos or --examples
options provided.

Also hide the classes tab and page in this case.

Bugfixes for class index and close tabs

For some reason Firefox 6 (and I haven't tested any other browser)
returns a.href attribute values URL encoded. For example:

  '#%21/api/OneClick.tld.GridList'

%21 is !

When encoded value is passed to handleUrlClick, it is not matched by
regexp, and javascript error is generated.

Better to have it master, than possibly out of sync.

Printing only works with --seo enabled.

With the require-rubygems section the file can be run from any
directory as it will know from where to load the jsduck library
files under Windows.

Using File.expand_path to expand all relative path to full paths,
which also converts all DOS-style paths to UNIX style.

Each member now has an :id property which is created in Merger.

Allows us to get rid of a bunch of duplicated and fragile code for
creating the ID from :static, :tagname and :name properties.

So normal and static members with same name will have different ID-s.

This fixes several problems with handling of static members.
Most noticably hide-inherited hides all inherited members and
filtering also hides the static member if they don't match.

Now comments like:

        /**
         Some text here
	 Another line.
	 */

are handled as the author intended.  Previously this text would have
been parsed by Markdown as code-block as it's indented.  Now we remove
the indentation also at the beginning of these lines that don't begin
with '*'.

Also include --builtin-classes option to short help message.

Added example of usage.