[Rpm-metadata] Testing on HEAD

[Paul Nasrat]

I think I've fixed up the relative/absolute/split path cases with the
new os.path.walk based finder.  I've nuked the unit tests for now but
I'd be gratefull if people could test out HEAD and see if it's working
from them in their use cases.

I know I need to make the cache code paths write to the correct
outputdir.  So if someone can clarify the expected behaviour I'll do
that.

Paul