Switch from Misaka to Python-Markdown

[fluffy-critter]

Context

Misaka is built on Hoedown, which is apparently abandoned, and has led to too much hacking on the Publ side anyway. Python-Markdown is much more modular and up-to-date, and is probably a better choice of Markdown engine. It also allows local customizations in a less-hacky way than the Misaka shoehornings, and being pure Python it also means one less thing that requires a C++ compiler on Windows builds.

In order to make this change, there's a lot of stuff that will have to happen:

• Ensure that the image spec parsing is close enough (in particular, supporting multiline expressions with spaces in them and the other "extensions" we use for image sets) — this is likely the hard part
• Same for link spec parsing
• Ensure that MathJax can still work, ideally with minimal change to the Markdown text
• Stop calling pygments ourselves, as Python-Markdown's built-in code-hilite module does that for us
• Probably a bunch of other non-obvious stuff

Also, I recall having evaluated Python-Markdown at some point and decided there was some reason I preferred Misaka at the time. Hopefully the issue I ran into is not actually an issue.

[fluffy-critter]

Looks like Misaka has been working on moving off of Hoedown. Hopefully this will fix the Hoedown-specific issues without breaking the parts of the API I've been relying on.