Investigate improving sha256 hashing

[cmwaters]

From Mus:

I believe we can also get another 3x in hashing performance (so maybe 27x in total) by simply changing the way we're hashing data to ensure that we're shoving as much data in the avx512 lanes as possible, something like this https://github.com/prysmaticlabs/gohashtree?tab=readme-ov-file

We should also investigate caching hashes wherever it makes sense as well (@evan-forbes)