Benchmarks

It's good to see this project is open source right now. Recently, I did some simple benchmarks for comparing `Marshal`/`Unmarshal` in different implementations and I also added fastpb. See https://github.com/rleungx/benchmarks/tree/main/protobuf. Please let me know if I was using the right way to do these benchmarks.

Start by reviewing the benchmark linked in the issue, particularly the protobuf directory, and verify how fastpb is used in the comparisons. Done requires a maintainer-confirmed methodology and a clearly identified follow-up, if any.

Written by the indexing model from the issue text.