Rewrite or refactor benchmarks using multicore-bench

[polytypic]

This PR changes the benchmarks to use the multicore-bench framework.

This PR also adds a bench.Dockerfile to run the benchmarks on OCaml 5.1 (rather than the default OCaml 5.0).

See also:

• Similar PR in Kcas.

---

It is clear that many of the data structure implementations currently suffer from false sharing and some other performance pitfalls. I briefly experimented with some optimizations (avoiding false sharing, avoiding indirections, avoiding contention by using a caching mechanism) and was able to e.g. get the spsc queue to perform more than order of magnitude better (from 15M/s to over 200M/s) on my M3 Max laptop. I'm, however, intentionally leaving the proper implementation of these optimizations for future work.

[lyrm]

Thanks for this PR. Multicore-bench is a way more readable framework than what we had before. That will make benchmarks more useful and hopefully less painful to write :)

Also, I like the ws_queue bench with Fibonacci. I will probably use this to write an example on how to use this queue.