This image file seems to have invalid utf-8 in xmp tags and getexif errors with

Traceback (most recent call last):
  File "/.../lib/python3.10/site-packages/PIL/Image.py", line 1542, in getexif
    xmp_tags = xmp_tags.decode("utf-8")
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xa8 in position 4312: invalid start byte

This is not technically a bug, but it may be better to print a warning instead of throwing an error. ImageNet is a popular dataset and this issue is going to be painful for its users.

The issue only happens with Pillow 11.2.1, because the previous versions do not have this UTF decoding step.
7d50816#diff-4805c79264fea07df59058db82ed74bb2f5c5023e212ac678536a534c56e5be2R1524