Caching pip is CI is buggy

[arunkannawadi]

I manually triggerred the end-to-end CI test, but it is still pulling an old version of roman_imsim_testdata submodule, which must be likely because of caching.

Log: https://github.com/DukeCosmology/roman_imsim/actions/runs/22426550261/job/64935922837

The hash of roman_imsim_testdata is setup is 5227befe71f6f690b5e9b9efbdc73b8f50b43d6c
when it should have actually setup f8e5b10f94b869fcd285480aa659703ec0b981de

I think we should drop caching, and replace bare pip with uv to speed things up a little.

cc: @sidneymau

[sidneymau]

Hmm, I would typically set things up using mamba. I've had some difficulty getting some things like esutil/etc. working correctly through uv installs (though I guess it has worked with pip so far... maybe the -Wno-incompatible-pointer-types thing fixed that)

[sidneymau]

Why would pip caching affect roman_imsim_testdata? It's not a package that gets installed and/or seen by pip? (It doesn't appear in the pip logs at all, from a quick glance)

[arunkannawadi]

False alarm. It was because the submodule was still pointing to the old commit. I pushed an update to it (directly on main)

[sidneymau]

Ah, I just did the same here 😅 9ac6b4c