Add Python version of the basic benchmarks

[abey79]

What

This PR adds the 3 basic benchmarks for Python. One notable different w.r.t the rust version is that the benchmarks use a single recording, but create a fresh memory sink for each iteration (the rust version creates a fresh recording for each iteration as well). This is due to #4410.

To run:

just py-bench

IMPORTANT: the python version of many_individual runs 100k points instead of 1M for the other benchmarks!

• Part of Report logging benchmarks for C++/Python/Rust to CI #4100

On my machine:

Checklist

• I have read and agree to Contributor Guide and the Code of Conduct
• I've included a screenshot or gif (if applicable)
• I have tested app.rerun.io (if applicable)
• The PR title and labels are set such as to maximize their usefulness for the next release's CHANGELOG

• PR Build Summary
• Docs preview
• Examples preview
• Recent benchmark results
• Wasm size tracking

[teh-cmc]

is there anyway to specify that the throughput of this benchmark is 100k per iteration, so that the stats make sense?

[abey79]

Good call. I've factored parametrisation in a better way that's explicit in the test name (and extensible if we want more test cases in the future).

Which reminds me to put a note that the num_point for many_individual is vastly smaller than the Rust version, as I actually want the benchmark to finish running inside of this year.