Encourage projects to add benchmarks.

[ericsnowcurrently]

With the recent pyperformance changes, projects can use it to run their own benchmarks. However, essentially no projects have such benchmarks written. Adding them would help projects avoid performance regressions (as long as they run them, of course).

It would also allow others (like us) to compose benchmark suites from across different projects. A sample of such suites could be used to train PGO or as the target for performance improvement. This would be particularly useful if the benchmarks were categorized (e.g. workload type, real-world vs. micro), via tags or something.

Concrete things we could do:

• post somewhere (targeting PyPI project maintainers) about the benefit of adding benchmarks and how to do it
• define/publish a GitHub action to run benchmarks
• publish a script to set up a benchmarking host/container (and/or publish an image), for projects (and GitHub) to use
• add tooling to discover projects with benchmarks on PyPI