Detect performance regressions via automated testing

It would be nice to be able to use Jenkins to more aggressively monitor for performance regressions.

There is currently one job wiredtiger-test-perf that runs and reports based on a small run of wtperf. We could extend that methodology and think about automatically triggering failures when regressions are detected.

I've created this issue so we can have a discussion about the workloads we can use that will most effectively detect regressions.