feat: improved `bit_rotl64` implementation for `keccakf` benchmark(s)

[DavePearce]

Summary

The various keccakf benchmarks use an implementation of bit_rotl64 which is suboptimal in terms of the number of ZkC instructions executed. Therefore, this issue should replace that implementation.

Approach

This is an improved implementation:

fn bit_rotl64(value:u64, n:u6) -> (res:u64) {
  var shift:u7 = 64 - (n as u7)
  res = (value << n) | (value >> shift)
}