Currently, if a command is run by benchRun that returns {ok: 0}, we increment the error count but don't actually print out the error. Instead, we should conditionally report the error if the op specifies showError: true.

This would make developing and debugging new performance tests easier.