Trigger selection

Implement dilepton trigger selection. Triggers that are prescaled in some portions of data, are only used in runs during which they were not prescaled. This is also mimicked in simulation. To do this, class RunSampler samples a run number for each event with a probability proportional to the luminosity collected during that run in data. The set of triggers to be applied is then chosen based on that run number.

Single-photon and non-isolated single-muon triggers discussed at the last meeting are included but commented out because they haven't been used when producing the current skims. We can enable those triggers after the next skimming campaign.

@usarica please check the lists of triggers.

Note that this approach with RunSampler can also be used in other places with potential time dependence. For example, we could have time-dependent pileup profiles, b-tagging scale factors, and so on. And everything would be correctly synchronized. For now I don't think we should go this way, but it's a nice option to keep in mind.