MAINT: Use FileNotFound rather than ValueError

[larsoner]

Opens a PR to address the comment in mne-tools/mne-python#9406 (comment), namely that trying to open a file that does not exist should raise FileNotFoundError rather than ValueError. In trying to test this, pytest failed for me, for two reasons:

1. The mfz file did not exist. Some sleuthing suggested that I needed to do python ./bin/mff2mfz.py ./examples/example_1.mff, but it seems cleaner just to do it automagically using pytest conftest.py, which is now implemented.
2. When I had some failures due to my changes, I'd get some failures not due to my changes because of poor test isolation in test_writer.py. Specfically, the use of .cache across a bunch of tests creates dependencies that can be problematic. Using the standard tmpdir pytest fixture is a clean, standard way to address this issue, so I implemented it.

I can revert these if you want, but in theory this should make it easier for your users (or CIs, if you want to implement them!) to do pytest and just have things work, and to iterate over tests more easily.

[ephathaway]

1. The mfz file did not exist. Some sleuthing suggested that I needed to do python ./bin/mff2mfz.py ./examples/example_1.mff, but it seems cleaner just to do it automagically using pytest conftest.py, which is now implemented.

Running make test will create the necessary .mfz file.

[larsoner]

Running make test will create the necessary .mfz file.

I think the autouse fixture is a bit more standard, do you agree it's worth keeping the new way? If so I can clean up the makefile not to need this and people can just do pytest (or make test) and have everything work properly. If not, I can delete the autouse fixture

[ephathaway]

do you agree it's worth keeping the new way?

Yeah, let's keep it. I think this will make it a bit more user friendly.

[ephathaway]

@larsoner, please take a look at larsoner#1. Thanks

[ephathaway]

@larsoner can you go ahead and remove the mff2mfz.py call from the makefile? Then I think this is ready to merge.

[larsoner]

I left the conversion target itself in there but removed the call to it in make test. Can remove the conversion target, too, if you prefer

[ephathaway]

I left the conversion target itself in there but removed the call to it in make test. Can remove the conversion target, too, if you prefer

I think the way you did it makes sense. Thanks. See larsoner#2 to address the GH workflow failure.

[larsoner]

The pesky "workflow awaiting approval" box needs to clicked to get CIs to run here

[ephathaway]

Thanks @larsoner!