Improve blake2b on arm performace

[shemnon]

Based on reference testing benchmarking the blake2b native code for ARM performs worse than the jave code running on arm. For this reason those forks of the build have been disabled.

We should try and use arm64 native means to improve the performance (be it compilers or custom vetor code fro arm silicon) and then re-enable the fork.

For darwin the gold standard are M1 and M2 macs.
For Linux ... ??? ... graviton instances?