Extract context patterns from webhook schemata

[woodruffw]

WIP. Still thinking through how I want to approach this.

Fixes #674.

Signed-off-by: William Woodruff william@yossarian.net

[woodruffw]

Noting for myself: doing a linear scan over thousands of fixed contexts for each dynamic context (i.e. parsed from expressions) is going to get slow, quick. This is probably going to require a prefix tree/trie structure so that I can do efficient prefix lookups.

Edit: Candidate: https://docs.rs/trie-rs/latest/trie_rs/#modules

OTOH a much simpler approach here is a regex/DFA transformation.

Related for me to look at: DAFSAs, DAWGs, etc.

https://en.wikipedia.org/wiki/Deterministic_acyclic_finite_state_automaton

[woodruffw]

Preliminary benchmarking shows a mixed effect on performance: running with --no-online-audits is about 9% slower on python/cpython, but about 6% faster on grafana/grafana.

(This is primarily a correctness change so even a small performance hit would be worth it, but it's good to see that it actually makes things faster in some cases.)