ATR-20586, Fix histograms filling from multiple threads

It turned out that in previous MR I made a mistake by removing the clone operation for HistogramFillers. That resulted in the fillers being actually shared by multiple "Group::fill" operations. Since these operations can occur in several threads at the same time the fillers variable were overwritten on the fly. An option would be to block filling for all variables and avoid cloning fillers. We can test in future what is faster. For now I have restored the solution in which locking is local to a single filler/histogram object which results in less contention.