[VisLab]

pandas 3.1 deprecates inplace= (Pandas4Warning, a DeprecationWarning) and pandas 4 removes it; under pandas 3.1 every validation with onsets warned and tests/test_deprecations.py failed.

• Remove the seven inplace= uses (df_util, key_map, data_util); ruff rule PD002 rejects new ones.
• Sort by onset with a stable sort, so rows with equal onsets keep their file order.
• Add data_util.remove_rows_by_value, which returns a new dataframe; deprecate delete_rows_by_column (removed in hedtools 2.0.0).
• KeyMap.remap fills "n/a" only into the target columns, and merge_dfs adds columns with .loc instead of .at (both pandas 3.1 deprecations).
• Tests: stable sort, the deprecation, the new function, and a class that runs each changed function with DeprecationWarning as an error.

[Copilot]

🔵 Needs a closer look

The version-sensitive pandas behavior still needs compatibility-test confirmation across supported releases.

No open issues 🎉

What changed in this PR

This PR updates dataframe utilities for pandas compatibility while preserving stable event ordering and introducing a non-mutating row-removal API.

Changes:

• Replace most inplace=True operations and enforce Ruff rule PD002, retaining an exception for the deprecated compatibility wrapper.
• Preserve file order for tied onsets and adjust target-column filling and schema-column assignment.
• Add regression coverage for row removal, deprecation warnings, and pandas compatibility.

File	Description
tests/​tools/​util/​test_data_util.py	Test non-mutating row removal and the deprecated API.
tests/​test_pandas3_compat.py	Cover stable sorting and warning-free dataframe operations.
tests/​test_deprecations.py	Track removal of the deprecated row-deletion API.
pyproject.toml	Enable Ruff's PD002 rule.
hed/​tools/​util/​data_util.py	Add non-mutating row removal and deprecate the in-place API.
hed/​tools/​analysis/​key_map.py	Replace in-place sorting and restrict missing-value filling to targets.
hed/​schema/​schema_io/​hed_id_util.py	Use .loc when copying columns into matching rows.
hed/​models/​df_util.py	Use stable onset sorting and replace in-place operations.

_{Review effort: Balanced}