DjangoUnicodeDecodeError for binary fields

[HealperRasmus]

Issue Description

The smart_str() function fails when handling binary fields, raising:
django.utils.encoding.DjangoUnicodeDecodeError: 'utf-8' codec can't decode byte 0xaf in position 2: invalid start byte

Steps to Reproduce

1. Create a model with a BinaryField
2. Enable audit logging for this model
3. Attempt to save/update the model with binary data

Current Behavior

Function fails with UnicodeDecodeError when trying to decode binary data as UTF-8

Expected Behavior

Binary data should be handled gracefully, possibly by converting to hex representation

Proposed Fix

The error occurs in the "smart_str" function. Adding a try catch with hex fallback could be a solution, but I cannot predict if that would have any unforeseen consequences?

def smart_str(s, encoding="utf-8", strings_only=False, errors="strict"):
    if isinstance(s, Promise):
        return s
    if isinstance(s, bytes):
        try:
            return force_str(s, encoding, strings_only, errors)
        except UnicodeDecodeError:
            return s.hex()
    return force_str(s, encoding, strings_only, errors)

[HealperRasmus]

Sorry, I just realized that smart_str is built in django function so not an option to edit.

The error happens here: auditlog.diff.get_field_value where it calls the the smart_str. So a better solution would be to handle it there:

Replace:

value = smart_str(
    getattr(obj, field.get_attname(), None), strings_only=True
)

with:

 value = getattr(obj, field.name, None)
if isinstance(field, models.BinaryField):
    # We cannot just use try/except here, because the binary fields might be memoryview objects, which
    # CAN be converted to string, but it will not actually represent the value stored in the database.
    # When larger bytes values are stored in the database, they are returned as memoryview objects.
    # in order to convert them to hex, we need to convert them to bytes first.
    if isinstance(value, (bytes, memoryview)):
        return getattr(obj, field.name, b'').hex()
return smart_str(value, strings_only=True)

An even more generic solution would be to define a dictionary of field handlers, mapping field types to functions that convert their field values to strings. This way its users have a way to handle their special cases without making a PR