FRI proof of work

[doutv]

We can refer to PoW mining Cuda code

[doutv]

Plonky2 FRI config pow bits = 16

We can keep FRI PoW in CPU:

1. CPU is fast enough. On a 32 cores server, it just takes 2ms, regardless of circuit size.
2. CPU PoW and GPU Commit & Query, they can be done in parallel. heterogeneous computing
3. PoW requires challenger after FRI Commit. challenger should be small, and it is fast to copy it from GPU to CPU