Avoid stale cache for multimodel statistics regression tests

[bouweandela]

Description

• Closes Multimodel statistics regression tests using sample data fail #1028

To show that this actually solves the problem, I

1. created a branch called fix-regression-multimodel-tests off master before Use nearest scheme to avoid interpolation errors with masked data in regression test #1021 was merged and pushed it, to run the tests and create the pytest cache.
2. pulled the current master into fix-regression-multimodel-tests and pushed that, to show that the tests fail because of the stale cache.
3. changed the cache key, committed and pushed to run the tests again and to show that this solves the problem.

You can view the results of the test runs here:
https://app.circleci.com/pipelines/github/ESMValGroup/ESMValCore?branch=fix-regression-multimodel-tests

I added some comments with instructions to change the cache key every time the cached input data is changed, to hopefully avoid this problem in the future.

---

Before you get started

• ☝ Create an issue to discuss what you are going to do

Checklist

• PR has a descriptive title for the changelog
• Labels are assigned so they can be used in the changelog
• Code follows the style guide
• Documentation is available for new functionality
• YAML files pass pre-commit or yamllint checks
• Circle/CI tests pass
• Codacy code quality checks pass
• Documentation builds successfully on readthedocs
• Unit tests are available

---

To help with the number pull requests:

• 🙏 We kindly ask you to review two other open pull requests in this repository

[valeriupredoi]

cheers @bouweandela - this was bothering me but I was under the assumption that'd be fixed by the periodic regeneration of the test container...until two days passed and the fail was still there 😁

[stefsmeets]

Looks good to me!

[stefsmeets]

Just a side comment, was this formatted by yapf? On my end it prefers the left side 🤔

[bouweandela]

It was formatted by me, yapf doesn't modify any strings (or any other part of the code). I feel that is more readable without the line-break in the middle of the argument.

[stefsmeets]

Agreed, but our autoformatters (if not yapf, then maybe docformatter) will prefer the left version so I gave up on changing it back 😅

[bouweandela]

I personally see those more as useful tools to quickly arrive at a decent formatting than the definitive guide on how stuff should be formatted. Hopefully a newer version of docformatter will be smarter.