Update diff commands in UMDP3 fixer

[Pierre-siddall]

UMDP3 fixer currently uses fcm diff and bdiff commands to take the difference between the equivalent of the main branch and the developers branch for example in this statement:

def get_branch_diff():
    """If in a local working copy of an FCM branch, return a
    list of files returned by running a branch-diff command"""

    # Use the bdiff command to extract a list of files that have changed
    # on the user's branch
    bdiff = subprocess.Popen(
        "fcm bdiff --summarize", stdout=subprocess.PIPE, shell=True
    )

    bdiff_stdout, _ = bdiff.communicate()
    bdiff_stdout = bdiff_stdout.decode(sys.stdout.encoding)

Therefore UMDP3 fixer needs to be refactored to reflect the move to GitHub.

[r-sharp]

After this was discussed in today's round table, I've taken a 60s look at the script.
Whilst we discussed a good few options, many of which I liked, to move this quickly on to something that runs as desired I think the simplest change would be as follows :

Hollow out the subroutine "get_branch_diff" and then use it to create a "Git_Bdiff" Instance from the github_scripts/git_bdiff module. (literally an import and a "variable = Class()" call.)
The 'files()' method in that object will return a list of all changed files.
Filter that list to create lists of Fortran and C files to return as the subroutine originally did.

That "should just work" tm

A path, with a default of "./" could be passed in to this new version as the init method of GitBdiff expects one.

Below is the simple filter method from the StyleChecker Class in umdp3_conformance.py, and whilst it expects the input to be a list of 'Path' objects rather than strings it's a reasonable example of how simple it is to filter the list into different ones.

    @staticmethod
    def filter_files(
        files: List[Path], file_extensions: Optional[Set[str]] = None
    ) -> List[Path]:
        """Filter files based on the checker's file extensions."""
        if not file_extensions:
            return files
        return [f for f in files if f.suffix in file_extensions]

Longer term, there is significant overlap in what umdp3_conformance and umdp3_fixer need to do to set up the loops over files, but not how the code is currently structured.
It looks like it would be easy to reuse sections of the Fixer within the checker (if for example everything insdide the do loop over 'files' were a callable method) then they could be imported to the checker quite simply and make use of the process parralelism it uses for checking files thus giving better performance on multiprocessor machines.
I think this should be discussed further and if the benefit is deemed worth the extra effort a plan to restructure both scripts around a common aim would be good.
I think fixing this issue as described above could be simple enough to do that anyway, while discussions on further work are ongoing.