When attempting to parse HTML rendered by Python Markdown, we would

crash any time an unknown named entity was encountered by

xml.dom.minidom. This has been found to happen when Python Markdown

processes an e-mail address that has a Unicode character in it that has

a known named HTML entity. Optimistically, Python Markdown will use this

instead of a character reference (&#...;), but this isn't known to the

XML parser.
Overriding the XML parser is fragile and error-prone, so it's not worth

doing. Instead, we simply undo what Python Markdown does by applying a

regex to convert all known named entities to character references.

These are safe to feed to the XML parser, and gives us the resulting

strings that we want.

Unit tests pass.