BUG: Raise TypeError when joining with non-DataFrame using 'on=' (GH#61434)

[iabhi4]

Closes GH#61434

What does this PR change?

When using DataFrame.join() with the on parameter, passing an invalid object like a dict, int, or third-party DataFrame previously resulted in unclear internal errors.

This PR adds a minimal type check that raises a clear TypeError when other is not a DataFrame, Series, or a list of such objects. Valid list-based joins without on remain unaffected.

Checklist

• Closes BUG: Joining Pandas with Polars dataframe produces fuzzy errormessage #61434
• Tests added and passed
• All code checks passed via pre-commit run --all-files
• Entry added in doc/source/whatsnew/v3.0.0.rst

[iabhi4]

All checks passed except the Pyodide build, which failed due to a rate-limit (HTTP 429). The failure seems unrelated to this PR. A rerun should resolve it

[rhshadrach]

This will raise when other is a tuple or other list-likes, whereas the current behavior can be successful. I'm -1 here.

[iabhi4]

This will raise when other is a tuple or other list-likes, whereas the current behavior can be successful. I'm -1 here.

Thanks for the feedback! I've updated the type check to avoid overly strict assumptions about other being a list, and it now handles generic iterables while preserving the original behavior. The tests have also been adjusted to reflect both invalid objects and mixed iterables.

Also, I wanted to check if this approach looks reasonable from a performance/complexity perspective.

The all() check on other is now applied only if it's an iterable (excluding strings/bytes), and I tried to keep it aligned with the original logic. Just wanted to confirm that the additional check won’t be a concern in terms of overhead for typical cases.

Open to adjustments if you think there’s a more efficient way to structure it!