Switch automated patch fetching on.

I asked Rok about this via e-mail just before the testing sprint but figure an issue is better since we can track problems once it's running if they arise.

Although there are test failures in core at the moment, it'd be extremely useful to have the TestBedBot fetching patches and checking just if they apply or not against HEAD (and only that) - then setting issues to Code Needs Work if they don't.

I'd estimate that out of the 300 or so patches CNR or RTBC, less than 100 actually apply. So we could weed out a very large number of stale patches before we even get to SimpleTests. This would give us human reviewers a massive headstart doing patch reviews, and will also mean there's a smaller number to keep up with when automated testing goes into full swing.