Refactor cellfinder workflow tests

[sfmig]

Suggestions for refactoring the cellfinder workflow tests, from the PR #23 review (thanks @alessandrofelder!)

• Can we redefine the integration tests as parametrised instead, considering a range of input configs (default, fetching from GIN, fetching locally)?
• Split the current integration tests into unit tests, each testing the individual steps (e.g., split setup into the 4 setup steps, and then test the actual workflow separately). This would allow us to remove the subprocess call and we think it would simplify the whole suite.
  • I would then use pytest's fixture caplog / capsys to assert logs
• Make the tests more modular.
  • Separate general unit tests (parser for now) and cellfinder-specific unit tests (with modules / directories)
• Start a utils module for all workflows (which for now would include the parser function)
  • (unit) test this separately
• Do we need the make_config helper functions?
  • if we are doing away with the subprocess call, probably no. Instead, have a fixture that returns the config as a CellfinderConfig class instead.
  • if we do need them, can we make them smaller? can they be defined inside the relevant fixtures instead?
• Integrate fixtures back into file with the tests' definitions, rather than in conftest.py
• Remove autouse=True for cellfinder_cache_dir fixture: this is because autouse=True will automatically execute the fixture for every test, and we won't want e.g. brainreg tests to trigger this fixture. (Can we have autouse only at module level?)

[alessandrofelder]

(Can we have autouse only at module level?)

Yes - we can define it in the module instead of conftest.py I think

if an autouse fixture is defined in a test module, all its test functions automatically use it. from pytest docs