Fix AssertionError when opening with chardet

[n-thumann]

This PR fixes an AssertationError, when chardet is unable to detect the encoding.

Added in 8d0d82b, it is checked if chardet has actually detected an encoding. If it's unable to detect one, an AssertationError is raised and codespell fails hard.

Example:

➜  ~ touch /tmp/empty_file
➜  ~ file /tmp/empty_file                 
/tmp/empty_file: empty
➜  ~ codespell /tmp/empty_file                                           
➜  ~ echo $?
0
➜  ~ codespell --hard-encoding-detection /tmp/empty_file                 
Traceback (most recent call last):
  File "/home/nthumann/.local/bin/codespell", line 8, in <module>
    sys.exit(_script_main())
  File "/home/nthumann/.local/lib/python3.10/site-packages/codespell_lib/_codespell.py", line 989, in _script_main
    return main(*sys.argv[1:])
  File "/home/nthumann/.local/lib/python3.10/site-packages/codespell_lib/_codespell.py", line 1178, in main
    bad_count += parse_file(
  File "/home/nthumann/.local/lib/python3.10/site-packages/codespell_lib/_codespell.py", line 871, in parse_file
    lines, encoding = file_opener.open(filename)
  File "/home/nthumann/.local/lib/python3.10/site-packages/codespell_lib/_codespell.py", line 218, in open
    return self.open_with_chardet(filename)
  File "/home/nthumann/.local/lib/python3.10/site-packages/codespell_lib/_codespell.py", line 231, in open_with_chardet
    assert encoding is not None
AssertionError
➜  ~ echo $?
1

With this patch we use always return f.encoding instead of encoding, which may be None, if chardet is unable to detect it. f.encoding either contains the encoding passed to open (if chardet detect it) or the encoding Python detected (if encoding is None).
Simply removing the assertation will make mypy unhappy, because the returned encoding can be None.

[DimitriPapadopoulos]

Excellent, we return io.TextIOBase.encoding, as actually set by open in text mode (checked it's UTF-8 for empty files).

[DimitriPapadopoulos]

The assertion was introduced by 8d0d82b / #2588 in an attempt to please mypy. That's not a proper fix, especially since Python can optimise assertions away, but mypy did give a good hint to look into this part of the code. And we now have the proper fix!