Speed up Subversion by only getting repo info when necessary.

Our Subversion support was looping over all server-side Subversion
repositories in order to compare UUIDs to find a matching repository.
This was pretty aggressive, and not really needed if paths matched.

We now do two passes. The first checks the paths. If we find a matching
repository, we're good. If not, we attempt to scan UUIDs, which is more
time-consuming, but hopefully not needed often.

This should speed up Subversion in the average case, especially when
there are lots of repositories configured for Review Board.

Ran against a local Subversion repository and reviews.reviewboard.org, and
with some extra debug info. Saw that it was checking the paths appropriately,
and fell back on UUID checks.