Stop Benchmark on SQLException for Testing

[apavlo]

Right now we will print but ignore any SQLException that occurs during a benchmark. This causes us to incorrectly mark benchmarks as passing in tests when they really did not. We should add a commandline flag that causes the benchmark to fail if any such error occurs.

[mariadb-DmitryVolkov]

Where can I vote for this issue? -) I guess this is pretty important!

[bpkroth]

#268 adds checks on the histogram outputs for benchmarks to make sure the errored transactions aren't "too bad".
Not exactly a "die early, die often" mode, but a step towards the spirit of this issue I think.

[bpkroth]

#427 also adds initial checks for the expected time and query count of tests

[bpkroth]

@apavlo do you think we need additional support for this to abort immediately on error, or is this good enough for now?