Zenodo too many requests failures in CI

[IAlibay]

CI is failing with 403 errors when pulling down from zenodo.

[atravitz]

As suggested by irfan earlier, I believe this is because our pooch cache is getting cleared. See here for an example where the pooch cache is not found: https://github.com/OpenFreeEnergy/openfe/actions/runs/20705516985/job/59435376352

[IAlibay]

At the cost of having mamba run longer, how about we remove the caching for the env building? Those are very large and are causing these issues.

I would also maybe suggest we drop some CI runners from our standard rotation (mostly python 3.11 & 3.12 macos).

[atravitz]

I'm in favor of dropping the python 3.11 and 3.12 macos runners, then see how the caches are.

I think it would be preferable to have the caches time out after ~2-3 hours, but I need to look into how to do this.

[atravitz]

short-term fix at #1820
longer-term fix (~2 weeks from now hopefully) is to move to cloudflare for hosting test data.