Run Pyrefly on pandas-stubs in CI

[yangdanny97]

Similar to #1240, this PR adds Pyrefly to run on just the stubs directory, since there are very few false-positives now (I've filed facebook/pyrefly#569 to address some of these).

I tested the CI and added the suppressions on my fork here: yangdanny97#1

There are some known gaps that prevent us from checking the tests directory without a lot of false-positives, so that can be saved for later.

[Dr-Irv]

One thing to consider. In CI, we have tests that separately test the stubs locally, but also test them when installed (for mypy and pyright). There have been occasions when the type checker passes the stubs locally, but fails when installed as a package. So you may want to add a test for pyrefly in that case, although maybe that doesn't make sense because you can't have pyrefly pass tests on user code (pandas-stubs/tests)

Let me know what you think about this.

[yangdanny97]

I was thinking that we test the installed stubs once we're able to check the tests directory. Without user code, I don't think we would emit errors for an installed dependency.

We're still some ways off from being able to do that (there are a few hundred false positives right now), so I figured we could run it on just the stubs for now and still get some value out of it.

Wdyt?

[Dr-Irv]

I was thinking that we test the installed stubs once we're able to check the tests directory. Without user code, I don't think we would emit errors for an installed dependency.

We're still some ways off from being able to do that (there are a few hundred false positives right now), so I figured we could run it on just the stubs for now and still get some value out of it.

Wdyt?

OK - that's fine by me.

[Dr-Irv]

thanks @yangdanny97

[Dr-Irv]

@yangdanny97 I'm going to assume that if new versions of pyrefly come out that affect these annotations, you'll update the version of pyrefly here and make any necessary changes to the stubs (e.g., you might end up removing a # pyrefly: ignore comment).