Describe the bug
I found that the Python API still requires a background channel even when skip_classification=True.
This seems wrong because if classification is skipped, background should not be mandatory.
Right now this blocks single-channel detection-only use cases.

To Reproduce

1. Create a valid 3D signal_array.
2. Call cellfinder.core.main.main(...) with:
   • signal_array=<3D array>
   • voxel_sizes=(5, 2, 2) (any valid voxel sizes)
   • skip_classification=True
   • do not pass background_array
3. It fails with:
   • TypeError: main() missing 1 required positional argument: 'background_array'

I also checked classification function behavior with background_array=None and saw:

• AttributeError: 'NoneType' object has no attribute 'ndim'

Expected behaviour
If skip_classification=True, run should work with signal-only input.
Background should only be required when classification is enabled.

Log file
No separate log file attached yet.
I reproduced using local script run:
python -u _bug_demo.py
and got the errors above consistently.

Screenshots
No screenshots (this report is based on Python/API repro, not UI testing).

Computer used (please complete the following information):

• OS: Windows
• Version: Windows 11

Additional context
This seems related to feature request #352 (make background channel optional).
I think this is a concrete blocker bug inside that larger feature work.