DEP: Remove tzdata as a hard dependency outside of Windows

[rhshadrach]

• closes BUG: pandas 3.0.0rc0 hard-requires tzdata Python package #63264 (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed.
• Added type annotations to new arguments/methods/functions.
• Added an entry in the latest doc/source/whatsnew/vX.X.X.rst file if fixing a bug or adding a new feature.
• If I used AI to develop this pull request, I prompted it to follow AGENTS.md.

This may still need more discussion.

Docs change:

[rhshadrach]

With tests seeming to be fine with no minimum here (ref: #63264 (comment)), I'm wondering if we should remove the lower bound.

[jbrockmendel]

No objection here. I didn't follow the conversation super-closely though.

[mroeschke]

Just noting that the conda-foge recipe should also be updated to make move python-tzdata to run_contrained (for windows). Could be done in conda-forge/pandas-feedstock#239 if for 3.0

[jorisvandenbossche]

or emscriptem?

[jorisvandenbossche]

I am fine with this, and also fine with removing the lower bound on the version.

Do you know what kind of error you now would get on Windows if you don't have tzdata installed? (I don't know if the import_optional_dependency you removed would have shown a warning before pointing to tzdata)

[Dr-Irv]

This probably will require a change to the pandas-feedstock for conda

[rhshadrach]

I added a test for the error message; a bit hacky, can remove if people prefer.

[jorisvandenbossche]

I see some other packages use sys_platform != "emscripten" instead. Do we know if there is a preference? (I don't directly find a good source about this in the docs)

[rhshadrach]

Interesting, I took it from here.

https://pyodide.org/en/0.16.1/faq.html#how-to-detect-that-code-is-run-with-pyodide

But looking at the latest version of the docs, it now uses sys.platform.

https://pyodide.org/en/stable/usage/faq.html#how-to-detect-that-code-is-run-with-pyodide

Will update.

[jorisvandenbossche]

Thanks @rhshadrach!

[jorisvandenbossche]

One comment on the raised error you now get when tzdata is not available:

zoneinfo._common.ZoneInfoNotFoundError: 'No time zone found with key Europe/Brussels'

That is the same error as you would get with an actual non-existing timezone (eg a typo). Would it still be useful to amend that message with some additional content if zoneinfo.available_timezones() gives an empty result? (just tested in pyodide, and there that function returns an empty set)

[rhshadrach]

No opposition assuming the implementation is reasonable; it's not clear to me what parts of the code we'd have to add try-excepts to do so. But I do think a more informative error message here is zoneinfo's responsibility (and would help all Python users, not just via pandas).