Fix px treating unsigned integer columns as categorical

[Belagum]

Link to issue

Closes #4291
Closes #4344

Description of change

_is_continuous() has a pandas fastpath that checks dtype.kind in "ifc", so unsigned integer columns (kind == "u") were treated as categorical. In px.bar that flips the auto-orientation heuristic, so a uint32 y column ends up drawn as thin horizontal bars and the plot looks empty. This adds "u" to the accepted kinds, as suggested in #4291 (comment) (the narwhals path already uses dtype.is_numeric(), which includes unsigned ints, so polars/pyarrow inputs were unaffected).

Demo

Running the example from #4291 with Rain_count cast to different dtypes, before and after:

input	before: _is_continuous / orientation	after
pandas uint32	False / h	True / v
pandas UInt32 (nullable)	False / h	True / v
pandas uint32[pyarrow]	False / h	True / v
pandas int32	True / v	True / v
polars UInt32	True / v	True / v
pyarrow uint32	True / v	True / v

Testing strategy

Added test_auto_orient_unsigned_int in test_px_input.py, parametrized over the existing constructor fixture (pandas, pandas nullable, pandas pyarrow-backed, polars, pyarrow) and UInt8/UInt16/UInt32/UInt64. It checks _is_continuous and the resulting px.bar orientation in both directions. Without the fix the 12 pandas cases fail, the polars/pyarrow ones pass. tests/test_optional/test_px passes locally (except test_render_mode, which needs statsmodels that I don't have installed).

Guidelines

• I have reviewed the pull request guidelines and the Code of Conduct and confirm that this PR follows them.
• I have added an entry to the changelog if needed (not required for documentation PRs).