revision 0 causes problems in SvnRepo.pdiff()

[OEP]

Calling pdiff() on revision 0 raises a CalledProcessError. This is sort of troublesome since it shows up as a parent of revision 1 and there's not a vcs-agnostic way of checking if all parents are valid inputs to pdiff().

I've got a branch over on my fork you can compare that seeks to remove revision 0 from the API if you like. It passes unit tests, though I don't think it removes the 0 revision in all usages of log(). At that, I'm not even sure removing revision 0 from the API is the right thing to do.

[ScottDuckworth]

I'm thinking the most graceful solution will be to return an empty diff.