reformat: complete remaining size_t index conversions

[metsw24-max]

Follow up f17110d by converting the remaining loop indices in avifImageYUVAnyToRGBAnySlow() from uint32_t to size_t.

This ensures pointer-offset multiplications involving row strides and pixel offsets are performed in size_t arithmetic rather than
potentially overflowing in 32-bit intermediates.

Changes

• Convert j and i loop counters to size_t
• Convert derived chroma indices uvJ and uvI to size_t
• Remove redundant cast-at-use stride multiplications

Rationale

The earlier sweep in f17110d standardized this approach across other YUV/RGB conversion paths. This patch applies the same fix
pattern to the remaining slow conversion path that was missed.

Testing

• Build succeeds without warnings
• Existing conversion paths remain unchanged functionally
• Large-image stride arithmetic now consistently uses size_t