Run integration targets in parallel
At present stem does integration tests in serial, starting a tor instance and testing against it one at a time. We should try running them in parallel, with a testing config option to keep the current serial behavior. This would involve using unique ports for each of the test instances, changing the get_runner() singleton in runner.py to return the right instance, watching out for mocking, and possibly other unforeseen gotchas. This might be a little tricky but if it works then the payoff would be sweeeet. Idea by Nick.