This PR #2367 by @orf refactors the benchmark suite to use Criterion. It also adds micro-benchmarks.

It would be neat if we:

• select a set of benchmarks
• automate their generation (maybe once a month).
• save the output (data or images) to https://github.com/RustPython/rustpython.github.io.
• create a webpage to show the results