Fix bug reported by @kcormier and better handle the processing of several samples interactively (in case of both data and MC being processed).
This goes together with the bugfixes to bamboo here (not merged yet): cp3-cms/bamboo!183 (merged)
The bug reported by @kcormier came from bamboo: for producing yields tables, bamboo creates special histograms, but these have no systematic variations produced (yet). However, there was already code in place for reading and using variation histograms (with __SYST suffix) if they were present in the output files, which is where the bug was present. Since we manually copy histograms produced in systematic variation samples into the nominal sample output file, adding a __SYST suffix, this triggered the crash.
The bug in bamboo is fixed, but we also now skip the copy of those special histograms when postprocessing alternative samples, since it doesn't make sense to include only those uncertainties in the yields tables.