Closes #901

Summary

While reviewing movement/utils/vector.py, I noticed a few places where NumPy operations were applied directly to xarray DataArray objects or where .values was accessed. These patterns can bypass xarray’s abstraction layer and may cause issues with lazy backends (e.g., dask) or metadata handling.

This PR improves compatibility with xarray best practices.

Changes

• Replace np.arctan2(cross, dot) with xr.apply_ufunc(np.arctan2, cross, dot) in compute_signed_angle_2d
• Replace in-place mutation of .values with xr.where
• Avoid direct .values access in cart2pol

Rationale

Using xarray-aware operations preserves metadata and ensures compatibility with lazy computation frameworks such as dask.

Testing

All existing tests pass locally:

pytest --ignore=tests/test_unit/test_napari_plugin

Result:
1011 passed, 4 deselected, 2 xfailed

No functional behavior of the library was changed; the modifications only improve implementation safety and compatibility with xarray.