ConfigParser chokes on conflict markers in stdin and throws AttributeError

Bug report

Bug description:

This is probably a duplicate of bug 107625, but I do not have a recent enough version of Python to check the behaviour.

Whether it is worth specially detecting conflict markers is something the code owners can decide.

To reproduce:

/tmp/py:

#!/usr/bin/python3

import configparser
import sys

sl = configparser.ConfigParser(allow_no_value=True)
sl.read_file(sys.stdin)

/tmp/z:

[list]
<<<<<<<< HEAD   (f12435 blah)
 foo
|||||||| BASE
 bar
========
 baz
>>>>>>>> CHANGE (faa768 blah)

$ </tmp/z /tmp/py
Traceback (most recent call last):
  File "/tmp/py", line 7, in <module>
    sl.read_file(sys.stdin)
  File "/usr/lib/python3.11/configparser.py", line 734, in read_file
    self._read(f, source)
  File "/usr/lib/python3.11/configparser.py", line 1077, in _read
    cursect[optname].append(value)
    ^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: 'NoneType' object has no attribute 'append'

Notes:

• sys.stdin seems to be needed to trigger this - deserves further investigation because it may point to a bad assumption or similar in the parser
• allow_no_value is needed to trigger this

Finger of blame:

            if (cursect is not None and optname and
                cur_indent_level > indent_level):
                cursect[optname].append(value)

This does not check that cursect[optname] is present.

CPython versions tested on:

3.10, 3.11, 3.12

Operating systems tested on:

Linux

Start in configparser.py at ConfigParser._read, especially the continuation handling around the reported cursect[optname].append(value) line. Reproduce the case with allow_no_value=True and conflict-marker input through stdin, then add regression coverage showing the parser no longer raises AttributeError and verify the existing behavior remains appropriate.