Switch filter and verify function

[lmas]

Currently using a Scalable bloom filter from github.com/tylertreat/BoomFilters but suspect it might not be performing optimally/correctly?

• Amount of hash functions doesn't seem to fit with optimal amount
• Amount of cells doesn't seem to fit with optimal amount (see above)
• Default hash function used results in inaccurate false probability rate
• Repo seemed to be a little messy in general: some models not documented, models missing common features found in other models, unhandled errors and use of unsafe (impressions from the time I wrote feedloggr v3 and first found BoomFilters)

Might want to try to implement a Stable Scalable Bloom filter myself instead:

• Stable Bloom won't keep growing forever (sort of deletes items from itself to stay with a fixed size)
• Easier to track, verify and document correctness
• Performance less relevant vs correctness
• One less unknown dependency

Or there might be another filter model that fits better? Some kind of Cuckoo filter?
Think I found the original paper for a scalable bloom filter though, doi:10.1016/j.ipl.2006.10.007

Investigate:

• https://github.com/bits-and-blooms/bloom

References:

• https://en.wikipedia.org/wiki/Bloom_filter
• https://ricardoanderegg.com/posts/understanding-bloom-filters/
• https://sagi.io/bloom-filters-for-the-perplexed/
• http://webdocs.cs.ualberta.ca/~drafiei/papers/DupDet06Sigmod.pdf
• https://bdupras.github.io/filter-tutorial/
• https://llimllib.github.io/bloomfilter-tutorial/
• https://www.waitingforcode.com/big-data-algorithms/stable-bloom-filter/read
• https://www.geeksforgeeks.org/bloom-filters-introduction-and-python-implementation/
• https://blog.cloudflare.com/when-bloom-filters-dont-bloom/
• https://news.ycombinator.com/item?id=33318788
• https://austinhenley.com/blog/challengingalgorithms.html seems to have some good examples? (also might want to save the article itself)
• https://gopiandcode.uk/logs/log-bloomfilters-debunked.html seems to have "the most correct formula" for calculating false positive rates?

[lmas]

As per 77853e2, stable bloom filter didn't work out at all. No idea why, but we're back with a scalable bloom filter instead.