Potential memory leakage because of storing verified captchas in-memory

[musq]

I see that we are storing already encountered hashes in memory (via self.verified_captchas = set()). I feel this could cause scaling issues because we only ever add items to that set but never remove it.

Maybe we can use an lru cache like data structure to handle automated garbage collection.

If it is not possible to handle this right away, I believe we should mention this as a caution in the README, so people are aware of it.

[cc-d]

Yes this is a good idea musq.

I'm currently banned from software engineering by ***** or I'd implement it.