Mwparserfromhell

Thanks to everyone for their patience with this release!

- Breaking change: dropped support for end-of-life Python 2.7 and 3.4.
- Added support for Python 3.8 and 3.9.
- Added binary wheels for Linux and macOS.
- Updated Wikicode.matches() to recognize underscores as being equivalent
to spaces. (216)
- Added a 'default' parameter to Template.get(), and implement dict-style item
access for template parameters. (252)
- Fixed a rare parsing bug involving deeply nested style tags. (224)
- Fixed parsing of section headings inside templates. (233)
- Updated HTML tag definitions.
- Internal refactoring and cleanup.

- Fixed an unlikely crash in the C tokenizer when interrupted while parsing
a heading.

- Fixed manual construction of Node objects, previously unsupported. (214)
- Fixed Wikicode transformation methods (replace(), remove(), etc.) when passed
an empty section as an argument. (212)
- Fixed the parser getting stuck inside malformed tables. (206)

- Dropped support for end-of-life Python versions 2.6, 3.2, 3.3. (199, 204)
- Fixed signals getting stuck inside the C tokenizer until parsing finishes,
in pathological cases. (206)
- Fixed <wbr> not being considered a single-only tag. (200)
- Fixed a C tokenizer crash on Python 3.7 when compiled with assertions. (208)
- Cleaned up some minor documentation issues. (207)

- Improved behavior when adding parameters to templates (via Template.add())
with poorly formatted whitespace conventions. (185)
- Fixed the parser getting stuck in deeply nested HTML tags with unclosed,
quoted attributes. (190)

- Added Wikicode.contains() to determine whether a Node or Wikicode object is
contained within another Wikicode object.
- Added Wikicode.get_ancestors() and Wikicode.get_parent() to find all
ancestors and the direct parent of a Node, respectively.
- Fixed a long-standing performance issue with deeply nested, invalid syntax
(issue 42). The parser should be much faster on certain complex pages. The
"max cycle" restriction has also been removed, so some situations where
templates at the end of a page were being skipped are now resolved.
- Made Template.remove(keep_field=True) behave more reasonably when the
parameter is already empty.
- Added the keep_template_params argument to Wikicode.strip_code(). If True,
then template parameters will be preserved in the output.
- Wikicode objects can now be pickled properly (fixed infinite recursion error
on incompletely-constructed StringMixIn subclasses).
- Fixed Wikicode.matches()'s behavior on iterables besides lists and tuples.
- Fixed len() sometimes raising ValueError on empty node lists.
- Fixed a rare parsing bug involving self-closing tags inside the attributes of
unpaired tags.
- Fixed release script after changes to PyPI.