feat(worker): bench setting the hash

[ppigazzini]

Raise worker version to 268 (also server side)

[ppigazzini]

The PR:

• benches for signature with only one process, using 1/5 of the previous time for a worker with 48 threads
• warmup the worker benching at depth 10
• benches using also the hash set in the test
• prints statistics only for the bench for nps
• prints the time used for the signature and the warmup

The nps bench for new_engine could be dropped to speedup further the process.

Note: alternate attempts to get a bench with lower Stdev (%), running consecutive benches to warmup and/or cooldown directly in run_parallel_benches, have failed with concurrency 48 at 1 thread. The CPU load was always at 100% as expected, courtesy of the warmup/cooldown benches. But max_nps / min_nps ~ 2, with 2x Stdev (%). Same result using only a cooldown bench, in order to have the to be benched processes starting at the same time.

def run_parallel_benches(engine, concurrency, threads, hash_size, depth):
    engine_list = [engine] * concurrency *3
    hash_list =[hash_size] * concurrency * 3
    threads_list = [threads] * concurrency * 3
    depth_list = [depth - 3] * concurrency + [depth] * concurrency + [depth - 3] * concurrency
    try:
        with ProcessPoolExecutor(max_workers=concurrency) as executor:
            results = list(
                executor.map(
                    run_single_bench,
                    engine_list,
                    hash_list,
                    threads_list,
                    depth_list,
                )
            )
    except Exception as e:
        message = f"Failed to run engine bench: {str(e)}"
        raise WorkerException(message, e=e)

    for result in results:
        print(result)

    return results[concurrency: -concurrency]