[Doc] tzdata error due to a lack of a discoverable system timezone database

[shr3yas-k]

Describe the enhancement requested

Some timezone-dependent PyArrow tests require a discoverable system timezone database (tzdata / zoneinfo).

On environments without standard filesystem layouts (e.g. non-FHS compliant systems or minimal containers), these tests can fail with errors like:

ArrowInvalid: Cannot locate or parse timezone 'CET' discover_tz_dir failed to find zoneinfo

There have been several mitigations over time like older interim fixes for Windows itself (#36996), PR #45425, and recent Windows CI fixes such as commit ab2c0ad.

Windows-specific failures appear to be largely handled now, but the underlying requirement and expected behavior are not documented for contributors on other platforms (e.g. non-FHS systems and hermetic environments).

Happy to follow up with a small documentation PR if this sounds reasonable. The goal would be to clarify that these failures are environment-related and expected in certain setups, rather than indicating a bug in Arrow itself.

Component(s)

Documentation

[rok]

@shr3yas-k did you want to improve existing documentation or add something new?

[shr3yas-k]

Yeah, I’d like to improve the docs by adding a short note explaining that this is environment related rather than a bug. I ran into it myself while testing locally on NixOS, so the idea is mainly to make this clearer for contributors on non-standard environments.

[rok]

Oh NixOS, interesting! Where are you installing pyarrow from?