Fix native floating-point frame decoding

[yinkev]

Closes #436.

Summary

• decode native FloatPixelData and DoubleFloatPixelData frames through ImageFileReader
• pass the detected pixel-data keyword into the shared frame decoder
• avoid requiring integer-only BitsStored and PixelRepresentation attributes for floating-point pixel modules
• add multi-frame float32 and float64 regression coverage, including negative fractional values

Root cause

ImageFileReader detected floating-point pixel tags while scanning the file, but read_frame() discarded that information. It then required integer Pixel Data attributes and called decode_frame() with its integer default keyword, so valid floating-point DICOM images failed before decoding.

Validation

• regression fails on upstream/master with missing BitsStored
• python -B -m pytest tests/test_io.py::TestImageFileReader::test_read_float_pixel_data -q
• full tests/test_io.py: 111 passed, 9 skipped
• repository-documented pytest --flake8: 2,022 passed, 178 expected optional-dependency skips
• python -B -m flake8 src/highdicom/io.py tests/test_io.py
• git diff --check

[CPBridge]

Thanks for catching this