Static hermes is slow. Document how to optimize

Here is the small sha512 implementation.

https://gist.github.com/paulmillr/f57268b66c3af9adf2423ef232cb0533

We've discussed the speed-up on twitter with @tmikov. The question is: how to make this much faster vs JS? Note that bigints aren't used internally - K512 is split into u32 arrays.

Otherwise there doesn't seem to be much of a point in the "blazing fast" Static hermes now