Rename Options::Parser to Options::HelpfulParser - this is the class
that works like Ruby built-in OptionParser, but with improved --help
system.

Use Options::Parser as a name for the class that constructs the
OptionParser object and performs parsing with it, returning
Options::Record object.  While doing so, it tries to do little more than
setting of attributes on Options::Record - it handles converting
filenames to canonical form; reading file contents for --head-html,
--body-html and --css; reading and interpreting the --config JSON
file (and also auto-detection of the latter).

Options::Processor takes care of the rest: expanding filenames,
rendering the footer, configuring various classes, and validating
the options.

So related things sit closer together.

For example previously the --encoding option triggered immediate
setting of Util::IO.encoding attribute.  Now we store it to @opts
object initially and then afterwards inspect this object to apply
the real effects of the options.

Similarly with several other options.

Storing the actual options in a separate OptionsRecord while processing
them inside Options class.

Lots of changes because a myriad of instance variable references needed
to be changed to point at @opts record.

Created an #def_attribute method for creating an accessor for an option
and setting the default value.

Now the options are initialized close to where they are modified,
making the logic easier to see.

Make this options also in line with other multiple-path-accepting
options.

So it behaves the same as other options that accept multiple paths.

Refs #216

Yet another thing that's not really a concern for the TagRegistry,
as it only effects a small subset of tags.

As a result, the TagRegistry class is now quite clean.

Again, not generic enough functionality to keep in TagRegistry.

Another responsibility that doesn't belong to TagRegistry.

TagRegistry doesn't need to deal with this stuff that's only used
in one place.  The Ext.define configs are only needed in Js::Class
and nowhere else.

Instead of prefixing each static method with "self.", use this
Ruby trick to place all these methods into the 'eigenclass'.

Add a generic #tags method to TagRegistry and do the filtering in
Web::ClassIcons instead.

To determine which icon to used when multiple possibilities.

For now primarily using this to choose the Singleton icon over
Component.

The @class_icon field is now a hash, containing three icons:

 * large - used in class header
 * small - used in tree, tabbar, search
 * redirect - used in search (for alternateClassName)

The Web::ClassIcons takes care of generating all the neccessary CSS.

Copy the files over to class_icons/ dir like the member icons
are copied to member_icons/.  Additionally the @class tag now
defines the default icon.

Similarly also generate CSS for the icons.

This change only deals with class icons in class header (icons in
tree and search are to do).

The ClassIcons class was separated already and this class doesn't
directly deal with icons.  Instead its primary purpose is to provide
the data for the class tree.

Use it in @component and @singleton to define a CSS class name for
the tag that's used for the class instead of the default.

Eliminate the hard-wired icon rendering of class header.

Though the CSS and icons themselves are still hard-wired.

Instead of checking for cls.inherits_from("Ext.Component") when
generating class icons, we can simply check if the class has the
component tag.

Process::Components takes care of auto-detecting components, so one
rarely needs to actually use @component tag.

Just doing things in more conventional way.

Don't assume that the params array is always present.  Even for methods.

Fixes #494

Convert latin1 escape sequences into UTF-8.  By default they resulted
in ASCII-8BIT (binary) encoding, which wasn't compatible with UTF-8,
resulting in error.

Added wrapper encoding method to cope with Ruby 1.8 which has no
builtin encodings support.

The tests file is now also a pure ASCII file, so no funky encoding-stuff
happens there.

Fixes #491

To handle the more complex nesting logic of the guide headings
and rendering of HTML, keeping the GuideToc class itself a lot cleaner.

It's a list, and should me marked up accoringly.

Turn GuideToc into instantiable class, so we can make use of
instance variables, refactor the logic into several helper methods and
avoid passing lots of arguments around.

Add tests for --guides-toc-level=1, 2, 3.

To be named more similarly with --guides option.

Improve the docs of the option - shorten lines and make the effect
of different argument values plain obvious.

The new `--guide-toc-level` option, which defaults to 2, controls what
heading levels will appear in guides' tables of contents. Specifying
`--guide-toc-level=N` will result in headings at or above <hN> being
included in all guides' TOCs. Choosing a `guide-toc-level` of 1 disables
the TOC entirely.