Fixing known test issues and small test refactoring.

[mattkwiecien]

Summary

Our tests don't consistently pass, which can be misleading for new contributors who find their new PR may lead to a failed build.

This was caused by 2 main issues, not completely tearing down the tests (e.g. cleanup) and a runtime error thrown by NaMaster that bubbles up the stack into our code.

With these changes I verified all tests pass (individually or all together) and all temporary files get cleaned up

NaMaster issue

• We call write_to on namaster workspace, covariances and this writes to a file. This can randomly throw a runtime exception "Unable to write to file"
• This error occurs randomly, about 20% of the time
• I tried locking and touching the file to make sure it exists, running synchronously, as well as writing to a separate temp directory. Nothing really changes the outcome.
• To solve this, I added pytest-retry which allows us to re-run tests that fail, I allow the tests affected by this bug to re-try 5 times. If we succeed once, we know it was the NaMaster issue and we pass the test.

Test changes

• I updated the tests to use setup_module() and teardown_module(). These run once at the beginning of each test file. This allows us to have more controlled reading/writing of files without constantly needing to check if files exist or deleting files that should still exist.
• Following typical naming conventions, I renamed any global scope variables in tests to be all caps
• I updated some of the mocks we use to be fixtures when possible for a more consistent testing approach
• All of the tests now write to tests/tmp/directory. Before they were writing to separate directories.
  • I originally had all tests writing to their own tmp/ subdirectory. This works but because of the code structure we would need 1 configuration file per test.
  • Instead of doing this now I left it as 1 directory, and if we ever run tests in parallel we can change this then.
• I moved all the configuration files for the tests to the same directory
• I removed any code that ran automatically at the global scope in the tests. This stops files from being arbitrarily written without actually running the tests.

Code Changes

• I added a clobber argument to get_covariance_block to allow a user to recalculate the covariance if they want to. Currently the only way to do that is manual or re-instantiate the class. This simplifies the test as well as giving more functionality to the user.

Build changes

• pip 23.1 officially removed support for --no-bin (see comments), this was causing mpich tests to fail. I updated the build script to use the new option --no-cache-dir

[mattkwiecien]

Note for myself

DEPRECATION: --no-binary currently disables reading from the cache of locally built wheels. In the future --no-binary will not influence the wheel cache. pip 23.1 will enforce this behaviour change. A possible replacement is to use the --no-cache-dir option. You can use the flag --use-feature=no-binary-enable-wheel-cache to test the upcoming behaviour. Discussion can be found at https://github.com/pypa/pip/issues/11453