Ensure that required tests always run.

[eap]

This change adds "notest" variants of existing required tests. The purpose of this change is to ensure that required tests always run without forcing them to consume resources with a full build when this would be unnecessary. While the main tests use the pull_request: ↳ paths-ignore: directive to trigger the workflow, this abbreviated test uses the same file patterns but applies the inclusive paths directive.

[climbfuji]

Do we really want this? I prefer to know that tests didn't run when I go back and look at the commits that were merged. This sort of fakes that tests were run when they actually weren't.

[eap]

@climbfuji

Yeah, that's a valid question. The problem we have is that we require these tests for merging so any PR that only includes files matching the paths-ignore filter will not be mergable without an override. GitHub doesn't give any guidance on the proper course of action here and the advice I've seen seems split between 3 choices

1. Don't require tests that don't always need to run.

2. Run required tests on every PR even if no affected code is changed.

3. Make a workflow with the same name that executes against excluded paths.

I think #3 isn't as viable as I originally thought, it would probably work in a binary case but this PR shows the weakness; in case you have a test that mixes paths then the workflow run is a race condition.

So... I don't know the best path here. I'm inclined against requiring tests that may not run but due to resource limits I really don't like activating all tests for no reason. We have plenty of PRs that only update docs.

[climbfuji]

In this case I'd say let's do (2), because we have the runners up and running 24/7 anyway. And the rebuilds from the buildcache take only a few minutes.

[eap]

moving the changes discussed here to #1210