[PATCH 0 of 4 STABLE] largefiles: fix various clone divergences from the unwrapped clone command

[Matt Harbison]

These issues where discovered when running the full test suite with '--extra-config-opt extensions.largefiles='.  It ends up eliminating a bunch of diffs across 40 or so *.t files, mostly along the lines of the wrong version is checked out on clone, which throws off subsequent unrelated output.

Note that patch 4 probably isn't suitable for stable because it wraps another function, but I included it here so the full picture can be seen and discussed at the same time.  I don't have a problem with changing that patch to follow the alternative mentioned in its commit message if that makes it suitable for stable, but it feels like caching the files should be a part of hg.clone(), not the command.  Copy/paste seemed like the least best option for it.