Monitor RQG code for performance regressions

If the RQG suffers a silent performance regression, this is going to impact the concurrency of the stress tests, making them unstressfull.

To avoid this, the RQG performance needs to be monitored via the Hudson farm by either:

- running DProf against gendata.pl and gensql.pl and storing the results and periodically reviewing them

- using a benchmarking/performance plugin from Hudson to store the performance data obtained by timing gendata.pl and gensql.pl runs.

We should monitor both the most complex optimizer grammar and the WL5004 grammar, because those are the queries most difficult to generate, and at the same time the workloads are different. It may be necessary to construct a single-threaded version of gentest.pl that does not use any Reporters, but does run the default Validators and other processing code.