Hash algorithms

Each hash algorithm has its factory method, on BloomFilter and on CountingBloomFilter:

Factory method	Hash	Width
`CreateXXHash128`	XXH128	128 bits
`CreateXXHash3`	XXH3	64 bits
`CreateXXHash64`	XXH64	64 bits
`CreateCrc64`	CRC-64	64 bits
`CreateXXHash32`	XXH32	32 bits
`CreateCrc32`	CRC-32	32 bits
`CreateAdler32`	Adler-32	32 bits

CreateXXHash128 is the fastest algorithm and the recommended default.

The hashes are not seeded: the same value always sets the same bits in a filter of the same size.

The limit of the 32-bit algorithms

A 32-bit algorithm cannot tell apart two values with the same 32-bit hash: whatever the size of the filter, the false positive rate cannot go below about itemCount / 2^32, about 2.3% for 100 million values. Use a 64-bit or 128-bit algorithm for large filters.

Avoid Adler-32

Adler-32 distributes short inputs poorly: many small integers or short strings share the same checksum. A filter that uses it has a much higher false positive rate than the one it was sized for, with sequential integers and short strings in particular. Use another algorithm unless you need Adler-32 specifically, for instance to match another system.