Apply lfs.fetchexclude filter to previous commits when pruning

[chrisd8088]

In PR #2851 the git lfs prune command was changed to respect the lfs.fetchexclude configuration option such that objects would always be pruned if they were referenced by files whose paths matched one of the patterns in the configuration option (unless they were referenced by an unpushed commit).

However, while this filter was applied to files referenced by recent refs (including HEAD), it was not applied to files referenced only by recent commits previous to a recent ref. This can result in the unexpected consequence that an object in HEAD is pruned because it is referenced only by a file that matches the lfs.fetchexclude filter, but a recent previous version of the object is not pruned despite also only being referenced by the same file.

We therefore add equivalent filtering to the logPreviousSHAs() internal function which is called by the ScanPreviousVersions() GitScanner method used by pruneTaskGetPreviousVersionsOfRef() in the git lfs prune command's main phase.

(Note that in PR #1743 the Git log scanning functions were refactored, and a common parseScannerLogOutput() internal function was created to be used when performing log scans during git lfs prune commands. This replaced the original logPreviousSHAs() function, which had support for path-based filtering; however, that functionality was apparently never used.)

We also add a test which confirms that git lfs prune respects the lfs.fetchexclude configuration option insofar as it prunes Git LFS objects for files whose paths match a pattern in the filter, both when they appear in commits directly referenced by a recent ref and when they only appear in commits previous to those.

Note that we explicitly use a gitattributes(5)-compatible form of pattern match (i.e., /foo/**) for the lfs.fetchexclude options in this new test because otherwise we will see failures as per #4945. When that issue is addressed in a future PR we will revise this test to use the /foo pattern match form in order to demonstrate that gitignore(5)-style matching is being performed.

Note too that we add one extra check to the existing "prune unreferenced and old" test for consistency with our new test.

/cc @larsxschneider as author of #2851.