fix(converter): validate scale factor finiteness and guard against below-1-pixel dimensions in ImageResizingConverter

[ThryLox]

This PR resolves issue #2162 by adding robust input validation to ImageResizingConverter.

Description

Currently, ImageResizingConverter allows positive non-finite scale factors (such as math.nan or math.inf). Furthermore, scaling an image with a valid positive scale factor can truncate dimensions to zero (e.g., scaling a 1x1 image by 0.5), which raises an unhelpful Pillow ValueError: height and width must be > 0 deep inside the processing pipeline.

Changes

• Imported math and added math.isfinite(scale_factor) checking to the ImageResizingConverter constructor.
• Added validation inside _apply_transform to check that the calculated dimensions (both new_width and new_height) are at least 1 pixel, raising a clean, informative ValueError before calling Pillow.
• Documented the raised exception in the _apply_transform docstring.
• Added unit tests to tests/unit/converter/test_image_resizing_converter.py covering:
  • Non-finite scale factors (nan, inf, -inf).
  • Scales that result in zero-size image dimensions.

[ThryLox]

@microsoft-github-policy-service agree