align_to_input_ext_scale=True still triggers Umeyama alignment, causing 2-view input failures

When True the predicted extrinsics are replaced with the input ones and the depth maps are rescaled to match their metric scale. When False the function returns the internally aligned poses computed via Umeyama alignment.

Expected Behavior:
When align_to_input_ext_scale=True and valid extrinsics/intrinsics are provided, the _align_to_input_extrinsics_intrinsics() call should be completely skipped, and the input parameters should be used directly.

Actual Behavior:
Regardless of the align_to_input_ext_scale value, as long as extrinsics/intrinsics are passed, _align_to_input_extrinsics_intrinsics() is always invoked, which internally calls align_poses_umeyama(), causing inevitable failures for 2-view inputs.

Locate the code path that handles align_to_input_ext_scale and calls _align_to_input_extrinsics_intrinsics(), then inspect how align_poses_umeyama() is reached. Verify the conditional behavior with valid extrinsics and intrinsics, including a 2-view input. Done means the alignment helper is skipped when the flag is True, while the existing internally aligned pose behavior remains when it is False.

Written by the indexing model from the issue text.