Implement a benchmark for zkEVM execution #6955

After reading this blog post by benbencik, I think we would find it useful to have a premade benchmark of real blocks and microbenchmarks, so we can optimize for the zkEVMs.

This has the benefit that, since the tests are deterministic, it's not vulnerable to noise and more friendly to autoresearch and similar.

The tool used by benbencik, ziskemu, should generalize decently to other zkEVMs.