utf-8-sig: UnicodeDecodeError offsets do not count the BOM #159071

Bug report

When the input starts with a BOM, the utf-8-sig codec reports UnicodeDecodeError offsets relative to the input without the BOM, i.e. error.start is off-by-3:

data = b"\xef\xbb\xbf\xff"  # Invalid byte at position 3
try:
    data.decode("utf-8-sig")
except UnicodeDecodeError as e:
    print(e)  # 'utf-8' codec can't decode byte 0xff in position 0: invalid start byte

The input is being sliced in utf_8_sig.py to remove the BOM, but raised exceptions are not modified to correct the offsets and object attributes. This occurs in a few places, e.g.

The utf-16 and utf-32 codecs also remove the BOM, but it is included in the offsets and object attributes in their errors, so this looks like an oversight.

CPython versions tested on:

3.14, 3.15rc2

Operating systems tested on:

Linux

Linked PRs

• gh-159071: Correct UnicodeDecodeError offsets in utf-8-sig codec when input starts with BOM #159081
• gh-159071: Fix utf-8-sig decode error offsets #159082