Set maximum version for fsspec because it is super unstable

[danielhfrank]

• When working on Update CI and switch to python 3.11 #57, I had to fix some behavior changes that crept in via an fsspec upgrade. In order to catch these issues in the future, I also set up a weekly test to catch any more regressions like that, which are easily possible since we don't pin dependencies in this library. Well, it's been two whole weeks, and fsspec released an updated version that somehow changed behavior again and showed up in the weekly test.
• I didn't really want to spend time chasing down what exactly changed, but I can see by toggling versions and running tests that something did. This is a silly problem to chase, so I put a version restriction that effectively keeps articat on fsspec==2024.2.0, which is what facets uses.

[dhimmel]

Wild. Are we using fsspec is a bespoke way or are they making frequent breaking changes to common functionality? (If this question takes more than 3 minutes to answer, please ignore 😄 ).

[dhimmel]

🟢 slight preference to comment with a link to this PR to save future editors a blame lookup.

[danielhfrank]

Done in 892ad57

[danielhfrank]

At least the last time that this came up, the functionality that was changing was around an fsspec implementation of glob, in a way that was attempting to work cross-platform. We make frequent use of the ** glob pattern, which is a little unusual I guess.

It would definitely have taken me more than 3 minutes to find out if that was in fact the problem again here, so I just pinned the version 😉