Performance comparison with Atomikos may loop forever

We have occasionally (though not common) seen our performance comparison (with other open source TMs) job loop forever when Atomikos fails to enlist a resource (for example job http://narayanaci1.eng.hst.ams2.redhat.com/job/narayana-performance/4/).

The correct behaviour should be to catch the error and abort the JMH run for that particular comparison.