Drop Python 3.10 support

[genedan]

We'll need to drop Python 3.10 support once this is done, since pandas 3.0 requires at least Python 3.11.

[henrydingliu]

Python 3.10 is not EOL for another 6 months. Do we want to drop it just for Pandas3?

[genedan]

Also responding to: #729 (comment)

Good point, I think we should keep 3.10 support until it reaches EOL. I think at a minimum, we need to have pandas>=2.3.3 in pyproject.toml. That's the lastest version of Pandas 2, works with Python 3.14, and is designed to prep users to transition to 3.0.

It's going to be a tough judgment call on when we change that to pandas>3.0. My vote is we do this at the latest when Python 3.15 gets released, since 3.14 is the last version supported by Pandas 2.

Another question is how to test Pandas 2 and 3 simultaneously without using too many resources. We can run every combination of Python 3.10-3.14 and pandas 2 and 3 or only the most common combinations. I'm going to push a commit that will update the workflow to test both pandas 2 and 3.

I think all the workflows run in parallel, so there shouldn't be a huge increase in execution time. But it does add to the complexity and you can opine when you see the output.

[henrydingliu]

It's going to be a tough judgment call on when we change that to pandas>3.0. My vote is we do this at the latest when Python 3.15 gets released, since 3.14 is the last version supported by Pandas 2.

This would be a brand new precedent, no? we have stuff like scikit-learn>1.4.2 and numba>0.54, etc. that are all extremely old but we are not applying this logic of "what's the latest version that's stated to support the newest version of python" to.