fix/update: corrected torch.py and added testing suite for utility functions

[kashtennyson]

Overview
This PR fixes a bug in the torch utility function and introduces a comprehensive unit testing suite for torch.py and image.py. This is a fix for #518

Changes

• utils/torch.py Fix:
  • Updated get_data_shape() to implement "passthrough" logic for non-tensor base cases (e.g., strings, labels, or None).
  • This prevents AttributeError when this function encounters metadata bundled with tensors.
• New Unit Tests:
  • tests/test_torch.py: Validates recursive operations on complex, nested, and mixed-type data structures.
  • tests/test_image.py: Verifies visualization helpers, including label formatting and annotator calls, using a robust mocking strategy.

Verification
The new test modules pass successfully:

poetry run pytest tests/test_torch.py tests/test_image.py

Results: 7 passed in 1.71s (4 in torch, 3 in image).

[dpascualhe]

Besides the pending comment, can you rename the test files to something like test_utils_image.py and test_utils_torch? Thanks!

[kashtennyson]

Renaming the test files to test_utils_torch.py and test_utils_image.py will break the naming consistency with the rest of the utility test scripts (e.g., test_conversion.py, test_io.py). Instead, I suggest organizing all utility test scripts under a new tests/utils/ directory. This will mirror the source structure while keeping the filenames clean and consistent. Please let me know your preference for this. Thanks!

[kashtennyson]

Hey @dpascualhe, still waiting on this PR's status and your preference on my suggestions in the previous comment. Thanks!

[dpascualhe]

Hi @kashtennyson , I like that suggestion! could you rename all test modules so that they mirror the directory structure within perceptionmetrics?