Investigate typing and mypyc

[masklinn]

Glyph makes a case for using mypyc in order to build binary wheels / accelerators (?) https://glyph.twistedmatrix.com/2022/04/you-should-compile-your-python-and-heres-why.html

Might be interesting to take a look at it, probably related to or after #93 as a replacement or improved API would be more likely to help with the typing necessary.

However:

• at the end of the day ua-parser is probably mostly dependent on regex performances (TODO: test), which mypyc likely can't do anything about
• glyph had to update their bench program to a "low level / optimised implementation" (in Python) to get sensible benefits, ua-parser is all about strings

[masklinn]

at the end of the day ua-parser is probably mostly dependent on regex performances (TODO: test), which mypyc likely can't do anything about

While not proven formally, I think the various benching and testing does over the years since this was open are pretty massive hints which don't require bothering with testing:

• cpython's regex engine is in c, mypyc can't accelerate that, only remove some of the overhead
• the JITted runtimes absolutely struggle with JIT-compiling the regex workload anyway
• linear matching is cooked pretty much by definition, Build & evaluate pure python regex filter #235 is almost certainly what we want to head towards rather than this
• mypyc could make sense on top of Build & evaluate pure python regex filter #235 but that seems somewhat unlikely, still could be worth a shot