RICH mirror alignment: optimize subset of mirror combinations, and HLT1 selection criteria (RICH-62, RICH-16)

To determine the RICH mirror misalignment compensations in the performance-wise optimal way, two key things are needed:

1. best mirror combinations subset as a maximum spanning tree of the bipartite graph with primary and secondary mirrors as two disjoint and independent sets of its nodes and with amounts of mutually reflected photons as their edges' weights;
2. for the HLT1 event selection line: precise solid angle selection windows for tracks that fill the photon"s delta theta vs. phi histograms for combinations from the above subset in the most efficient way.

This stores the list of the best RICH2 mirror combinations subset in Moore/-/blob/master/Hlt/RecoConf/python/RecoConf/rich_data_monitoring.py.