TST: add benchmarking and Faker support

[aqw]

The core motivation for this PR is to add some benchmarks and test out CodSpeed to see if it's a reasonable fit for us.

If not, the benchmarks are still useful, but we'll go back to vanilla pytest-benchmark. Codspeed is mostly compatible, until it isn't. For example, it does not support pendantic, which is a loss.

Generating large, representative repos led me to finally add Faker support. I hope/plan to use this for more tests that require populated repos, rather than our current verbose, hard-coded mess.

This PR also:

• fixes a bug in git.get_hexsha()
• fixes a documentation bug WRT flake8->ruff
• normalizes onyo_cat()'s arguments to match other onyo functions.

Fixes:

• 224

• 254

I encountered numerous bugs around pytest fixture scoping, parametrization, indirect parametrization, and referring to the parent class.

As a result, I hack a class-level cache through a function-scoped fixture. I would like to to this differently, with a class-level fixture populating and function-level fixture cloning, but this was not to be.

I also setup Faker with fixtures for each scope. We may not use this, in which case we can remove it. But I wanted to include it for now so we have a history of it in the repo, and it may in fact turn out to be of use.

[aqw]

I have pushed a new set of commits to this PR.

Primary changes are:

• faker is now available across multiple scopes
• drop CodSpeed and use pytest-benchmark directly (soooo much nicer)
• Add a class-scoped ui marker

If there are any questions on my approach, I am happy to explain. I'm not thrilled with the solution, but it was the best I could manage given the numerous pytest bugs I encountered.

[aqw]

Thanks for the review!