DCA performance test.

[gbalduzz]

If @biddisco managed to integrate a performance metric in the testing suite it might be a good idea to develop a performance test of the whole DCA algorithm. My question is: should we focus on the Monte Carlo only, or use a full DCA loop? Should we use DCA or DCA+? Input from @maierta is especially welcome. We could also have one MC only and a separate DCA loop test. How big should be the system? I suppose using something as big as the recent manual benchmarks is not feasible. Can we use an expansion order between 100 and 500? How long should this single node test(s) take?