CI: Install and test with Python and fslavwutils

[effigies]

Gets tests working from Python 3.6-3.11. 2.7 clashes with fslavwutils, which prevents testing while there's an fsl dependency.

It looks like there's not a lot of testing going on (coverage is 5%)... Do you have test data you use to actually run the program? I know there are some proprietary files we probably can't publish, but presumably we could make some identity file just to make sure the code is exercised and changes don't damage that.

[coalsont]

I don't think we have any fake coefficient files at the moment, and the real ones are considered proprietary manufacturer info, yes.

[glasserm]

Does anyone understand the coefficient file enough to know how to make an identity coefficient file for the perfect scanner?

[effigies]

It turned out to be trivial. I haven't looked through all of the coefficient file formats, but at least for Siemens, the code is creating zero arrays and then updating them from the files. By creating an empty file, we use all-zero coefficients, and it turns out that this produces unwarped files that are unchanged from the source.

This at least gets coverage up to 64%.

I think this is a good place to start. It won't replace manual checks with vendor files, but should at least make it obvious if a refactor breaks things.