Test organization improvements

[MeGaGiGaGon]

Part of #2238 While working on my first PR to the repo, I've noticed that interacting with the test cases is not easy. This is because of how much is in each file, with the longest being ~1200 lines long. Because it's a test that uses # output, the input and output for a specific case are ~600 lines apart. This is made worse by the fact that formatting changes line counts, so getting back to the source isn't as easy as dividing the line number by 2. Running the test case also doesn't help, as it only reports the end failure, and nothing about the original input.

With the current framework, the way I can see to solve this would be splitting cases into their own files, inside folders. For example, instead of one big preview_long_strings__regression.py file, there would be a preview_long_strings__regression folder with contents named preview_long_strings__regression_1.py, preview_long_strings__regression_2.py, etc.

This would be really nice for making all the tests more local, as if a set of tests fails, each one would have it's input directly next to it's output, and each test case would be individually re-runnable with pytest.

I'm not sure how this would affect the performance, since I don't know the cost of running a bunch of tiny files vs one large file.

I'm not sure if the current test runner supports running cases inside folders, but given the current structure I assume so.

If for whatever reason a test ever needed to be removed, this would make it a different type of difficult, as instead of having to find the input in a long file, now all test files after the one being removed would have to be renamed.

If this is a thing that is wanted, I'd be happy to start work on the PR.