PaDim occasionally NaNs in anomaly map

[VdLMV]

With custom dataset and backbone distances sometimes contains small negative numbers, possible due to imperfect pseudo inverse calculation. Solution was to clip distances to zero before sqrt.

• Solution is to clip distances to zero before sqrt.

• Fixes # (issue)

Changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• This change requires a documentation update

Checklist

• My code follows the [pre-commit style and check guidelines](https://openvinotoolkit.github .io/anomalib/guides/using_pre_commit.html#pre-commit-hooks) of this project.
• I have performed a self-review of my code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing tests pass locally with my changes

[VdLMV]

@samet-akcay : I don't understand pre-merge fail. Plz your help.

[samet-akcay]

@VdLMV, looks like some of the tests in tests/pre_merge/deploy/test_inferencer.py fails with this change. I'll check it out and let you know.

[samet-akcay]

@VdLMV, your change is fine, I'll approve it. However, PADIM started to fail in the tests for some reason. It fails when the model is exported to ONNX. We'll investigate this a bit deeper.

[samet-akcay]

@VdLMV, the reason turned out to be the latest torchvision version, which causes some issues on the ONNX / OpenVINO side. I'll merge this one now.