Add reproducibility test for slices running alone vs when part of full menu

New test in which the full menu is run first followed by tests of specific slices. All on the same events. Next the logs are compared if the decision counts (the count of objects rather than the count of events differ). Presence of the difference signifies cross-talk between slices. At the moment muon and bphys have a different result. Tagging @rbielski @bwynne @sshaw