Gexml

* Handle list tagnames, and field tagnames within a list.

* Implemented a basic project website, with beginnings of tutorial
and API docs.
* Implement Model.render(pretty=True) for pretty-printed XML output.
* Fix rendering of unicode String fields; thanks saltycrane.
* Clean up and clarify handling of wrapper tags for List/Dict fields.
When the "tagname" property is specified:
* If the field is required then the wrapper tag is required,
even if it has no contents.
* If the field is required then the wrapper tag is always rendered,
even if it has no contents.
* If the field is not required then the wrapper tag can be elided,
and will not be rendered when the collection is empty.
Behaviour of List/Dict without a wrapper tag is unchanged.

* Got 100% test coverage on all target python versions. Many edge-case
bugs squashed in the process.
* Eliminate all use of "raise RuntimeError"; use ValueError instead
because they were all about incorrect values encountered at runtime.
* Inherit meta-attributes from base classes in left-to-right order.
* fields.Boolean: coerce all values to a proper bool.

* Fix handling of whitespace around List/Dict fields; thanks Alexey Luchko.

* Some API changes to allow generator-based rendering:
* Model.irender() is a generator version of Model.render()
* Model.render() no longer takes the "nsmap" argument
* fields.Model no longer calls Model.render(), so reimplementing it in
a subclass probably won't work; reimplement Model._render() instead.
* fields.List accepts a generator, and won't consume it until rendered.
* fields.Model: when specifying the model class as a string, search
first by tagname and then by classname.
* fields.Value: allow a tagname of "." to parse a value directly out of
the contents of the parent node.

* Automatically escape/unescape special characters in string fields.
* Parse CDATA_SECTION nodes as if they were TEXT nodes.
* Added CDATA field type, which is just like a String except it renders
as a CDATA_SECTION node.