Transition away from Flake8 for linting

[jack-fs]

Linting in CI has broken one too many times. Maybe we should replace flake8

• Look into flake8 problem
• Investigate alternatives
• Determine configuration
• Update code if need be (hopefully minor)
• Update GitHub actions

[jack-fs]

Useful (though outdated?) overview:
https://pawamoy.github.io/posts/python-static-code-analysis-tools/

Tool	Notes	Pros	Cons
prospector	wraps PyFlakes, pycodestyle, McCabe, plus lots of extensions	Aggregator: a superset of flake8 in spirit	No extension ecosystem (is this a con?)
	Similar in idea to flake8: wraps pycodestyle, McCabe, but uses PyLint rather than PyFlakes by default	Explicitly aims to be less noisy/confusing
	Has additional features, broader scope compared to flake8 but no extension ecosystem
pylama	Wraps pylint, mypy and others	Extensible	Seems strictly less compelling than prospector, based on activity and feature set
Ruff	Implemented in rust	Extremely fast	New
	Intended to work alongside autoformatters, so stylistic lint rules are de-emphasised	intended as flake8 replacement	Under very active development
		reimplements flake8 features	Only a subset of flake8 lint rules
			Editor/LSP integration is WIP
Related tools
PyLint	More complex checks/suggestions than flake8		Slower than flake8
pycodestyle	pep8 checker
mypypyrightpyre/pytype	static type checkers
bandit	security-focussed; not particularly applicable for our project

[jack-fs]

tholo/pytest-flake8#87 and its discussion seems to be relevant, given the errors being thrown in CI. One recommended workflow is to run linting/formatting as a pre-commit hook, rather than in CI. This would require us to coordinate to use pre-commit hooks. I don't think this is necessarily an either/or scenario at our scale; and kind of doesn't answer the central problem: the errors being thrown by pytest

[dsteinberg]

This is failing again - this time with a different problem.....

https://github.com/gradientinstitute/causal-inspection/actions/runs/3545436523/jobs/5953583059

[jack-fs]

Seems to be due to flake8-quotes: see zheller/flake8-quotes#115. The new version 3.3.2 seems to fix it: see the current CI run https://github.com/gradientinstitute/causal-inspection/actions/runs/3879072775