Add option to preserve additional columns when loading from txt

Summary

Add a parameter to load_trajectory_from_txt that preserves columns beyond the standard id, frame, x, y instead of silently dropping them.

Background / Context

The txt loader currently uses usecols=[0, 1, 2, 3] in pd.read_csv, discarding any columns past the first four. Users who have extra data in their txt files (e.g. z, group id, or other per-agent attributes) lose it on load. Constructing TrajectoryData directly from a DataFrame already preserves extra columns, so the file-based loader is inconsistent with the direct construction path.

Technical Details

• Column selection happens in _load_trajectory_data_from_txt via usecols=[0, 1, 2, 3] and names=[ID_COL, FRAME_COL, X_COL, Y_COL]
• The txt format has no column headers, so extra columns beyond the standard four have no names — the parameter should accept an optional mapping or list of names for those columns
• TrajectoryData.__post_init__ already passes extra columns through unchanged, so no changes needed there

Acceptance Criteria

• load_trajectory_from_txt accepts a new optional parameter to preserve additional columns
• Extra columns are accessible on traj.data after loading
• When the parameter is not set, existing behaviour is unchanged (extra columns are dropped)
• New behaviour is covered by tests
• New parameter is documented in the API reference

Start with load_trajectory_from_txt and _load_trajectory_data_from_txt, especially the pd.read_csv usecols and names arguments described in the issue. Add tests for preserving named extra columns and for unchanged default behavior, then update the API reference; done means extra values are available on traj.data while the default still drops them.