grass.gunittest: Fix parsing exclusion from config file on Windows

[echoix]

Description

Refactored the function fnmatch_exclude_with_base in grass.gunittest.loader in order to work across platforms, including Windows.

I tried multiple approaches and researched a bit, and the first really working solution was the custom fnmatch_ex function that pytest uses, since pathlib's Path.match() doesn't work properly and doesn't behave like Path.glob() concerning **s.
https://github.com/pytest-dev/pytest/blob/9515dfa58a144f3644fd29b256113d723c9c1955/src/_pytest/pathlib.py#L414-L447

def fnmatch_ex(pattern: str, path: str | os.PathLike[str]) -> bool:
    """A port of FNMatcher from py.path.common which works with PurePath() instances.

    The difference between this algorithm and PurePath.match() is that the
    latter matches "**" glob expressions for each part of the path, while
    this algorithm uses the whole path instead.

    For example:
        "tests/foo/bar/doc/test_foo.py" matches pattern "tests/**/doc/test*.py"
        with this algorithm, but not with PurePath.match().

    This algorithm was ported to keep backward-compatibility with existing
    settings which assume paths match according this logic.

    References:
    * https://bugs.python.org/issue29249
    * https://bugs.python.org/issue34731
    """
    path = PurePath(path)
    iswin32 = sys.platform.startswith("win")

    if iswin32 and sep not in pattern and posix_sep in pattern:
        # Running on Windows, the pattern has no Windows path separators,
        # and the pattern has one or more Posix path separators. Replace
        # the Posix path separators with the Windows path separator.
        pattern = pattern.replace(posix_sep, sep)

    if sep not in pattern:
        name = path.name
    else:
        name = str(path)
        if path.is_absolute() and not os.path.isabs(pattern):
            pattern = f"*{os.sep}{pattern}"
    return fnmatch.fnmatch(name, pattern)

However, after inlining more and more of that fnmatch_ex function, comparing how it behaves with our inputs (Windows and Linux) and what branches are taken, we don't use a mix of absolute and relative paths, and we never supported **, and we never only supplied a name without separator. So, since on Windows the working fnmatch_ex needs a pattern with backslashes \ for the path and the pattern, and on Linux it needs forward slash on both; since pathlib's Path representation already normalizes a ./ of a relative path; and since pathlib's Path/PurePath classes works properly with * and **, especially when that variable is converted to a string at the end, I could simply replace all our logic by creating Path instances for the full name and pattern, and call the fnmatch with the string representations of it. I used PurePath instead of a concrete Path, as I didn't need to actually manipulate files, and the subclasses can be instantiated on any OS if further tests would need to run cross-platform.

Also included:

• the files list variable in discover_modules could not exist in all code paths, depending on the inputs. Initialized it with a list containing the same elements as all_files, without directly assigning all_files to files.

Motivation and context

Fixes #4319

How has this been tested?

Up until my final clean rebase, I was using extra prints inside the function, to not have to wait at the end of the test run to see if the files were properly excluded.

    print(f"Ed: fnmatch_exclude_with_base_pathlib, exclude: {exclude}")
    print(f"Ed: fnmatch_exclude_with_base_pathlib, patterns: {patterns}")
    print(f"Ed: fnmatch_exclude_with_base_pathlib, base: {base}")
    print(f"Ed: fnmatch_exclude_with_base_pathlib, base_path: {base_path}")
    print(f"Ed: fnmatch_exclude_with_base_pathlib, files: {files}")
    print(f"Ed: fnmatch_exclude_with_base_pathlib, not_excluded: {not_excluded}")

Screenshots (if appropriate)

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing
  functionality to not work as before)

Checklist

• PR title provides summary of the changes and starts with one of the
  pre-defined prefixes

• My code follows the code style
  of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added tests to cover my changes.

[wenzeslaus]

Using pathlib is definitively better!

[wenzeslaus]

I really meant to request changes, not approve.