Collect RR and DIR results using collect_results.py

So far, collect_results.py can only handle verification modes, but no closed- or open-set recognition. This MR adds the possibility to collect results for the rank 1 recognition rate (RR) and the rank 1, threshold@FAR=0.001 detection and identification rate (DIR).

So far, parameters like rank or FAR threshold are not specifiable on the command line. They have not been specifiable for verification either. If wanted, I can add these parameters to the command line options.