Mf2py

- Stricter checks that Parser.__init__ params are actually None before
ignoring them.

- Now produces unicode strings for every key and value, no more byte
strings anywhere.
- Do not add 'T' between date and time when normalizing dates
- Unit tests for running the microformats test suite

- New top-level "rel-urls" entry, contains rich data parsed from rel
links, organized by URL.

- convenience method `mf2py.parse` that takes the same arguments as Parser
and returns a dict.
- nested h-* classes now parse their "value" based on the property
they represent (p-*, u-*, dt-*), so for example "p-in-reply-to
h-cite" would have a name as its value and "u-in-reply-to h-cite"
will have a URL.

- Add rel=bookmark to backward compat parsing rules based (translated
to u-url in mf2)
- Parser constructor now takes explicit named arguments instead of
**kwargs, for saner behavior when called with unnamed arguments.
- Bugfix: Empty href="" attributes are now properly interpreted as
the current document's URL.

- Minor Py3 compatibility fix
- Correct typo `test_requires` -> `tests_require` in setup.py