Improve scivision testing

[edwardchalstrey1]

It would be worth taking a look at the various functions/modules in scivision/catalog and scivision/io and adding some more tests to ensure everything works as expected.

pytest is already set up, so just add tests to /tests

• Tests for functions in:
  • scivision/io/reader.py
  • scivision/io/wrapper.py
  • scivision/io/installer.py
  • scivision/io/autoplumber.py
  • scivision/catalog/catalog.py
  • scivision/catalog/base_catalog.py
  • scivision/koala.py
• Update GitHub url parsing to use an external library if one exists (especially if there's one that works for other sites e.g. bitbucket) <- EC: doesn't seem like one exists
• Make sure that tests which rely on existence of external repos are controlled via specific commit hash
• Use e.g. papermill to run the example notebooks as part of the automated testing (to check that they don't fail - not necessarily to validate their output)
• Update binder to use themain branch version of scivision when everything is tested properly - currently we have a temporary fix of using the pypi version of scivision, to avoid breaking changes pushed to main - but in future tests should prevent this - undo Replace postbuild with the pip scivision version #145
• Update bad test in test_catalog.py #179
• Modify any tests that rely on remote urls so they don't
• Modify any tests that involve installing extra packages/ dependencies, such as the tests for load_pretrained_model which install a scivision test plugin model that has tensorflow as a dependency and prevents Python 3.11 compatibility (e.g. instead load a model with no dependencies that has a single know return value for its predict func)

[edwardchalstrey1]

@kasra-hosseini @ots22 I've updated the testing infrastructure a bit as you'll have seen in recent PRs, I think I could use some help in a group session at some point to understand the remaining un-tested modules in the above list better - but I'm going to bump this issue down in the project board for now as I don't think it's urgent

[kasra-hosseini]

Great, thanks @edwardchalstrey1!

[edwardchalstrey1]

Use mocking in tests instead of interacting with remote urls: https://alan-turing-institute.github.io/rse-course/html/module05_testing_your_code/05_04_mocking.html#using-mocks-to-model-test-resources