Build & evaluate pure python regex filter
FilteredRE2 (and regex-filtered) show amazing results on ua-parser work load, however if for some reason they're not available (or they're missed) users are stuck with the horror show that is sequential regex matching.
regex filtering should be reimplemented for the stdlib's re module in order to see how it behaves and if the gain is as significant as it is for the native regexes. The big question will be how efficient a pure python aho-corasick is (possibly via ahocorapy?)
See prior art at https://github.com/ua-parser/uap-rust/tree/main/regex-filtered
Do take note of ua-parser/uap-rust@4f1c7df94f4171d78e0128146254d99e6a12c27e and ua-parser/uap-rust@f500c57dbc100d1845ca73d7253240bf258994bb
An other core concern will be the implementation of aho-corasick in Python, uap-rust was able to use https://github.com/BurntSushi/aho-corasick but a third-party dependency would be a bigger concern here.