This introduces DiffXParser, a new diff parser that can parse DiffX

files, using the pydiffx module. SCMTools can provide this as the diff

parser if they permit DiffX files to be used (whether DiffX is used

natively, or if they look for a #diffx: header in get_parser()).
This parser first turns DiffX file contents into a series of DOM (DiffX

Object Model) objects, which are then turned into ParsedDiff objects.

The ParsedDiff objects are populated with all the information needed

to reconstruct the DiffX file perfectly (options, preambles, metadata,

and diff contents).
That reconstruction happens when downloading a raw diff. With normal

diff parsers, that reconstruction just builds up a byte string

containing the diff content for each file (which may include metadata or

special lines for that diff variant).
With DiffXParser, the content is instead built up using that parsed

information stored in extra_data, along with the actual diff content

(that in a #...diff: section). It's assembled back into a DOM, and

then serialized.
All the extracted data is also exposed to the API, SCMTool

implementations, and extensions. It's all stored in namespaces within a

diffx key in the extra_data fields in DiffSet (for the main DiffX

section), DiffCommit (for change sections), and FileDiff (for file

sections).
Along with this, a diff's encoding is extracted and stored in a

top-level encoding key in FileDiff.extra_data (which is an existing

established key that aids in diff processing).

Unit tests pass on Python 2 and 3.