Preliminary fixes for a new analysis

This MR is preliminary to the addition of a new analysis using a Multi Variate approach. It consists in a couple of changes, potentially setting the ground for a more general fix:

• The line added in AnalysisClass.h allows the possibility to have multiple BDTs (not TMVA)
• The rest of the changes allows to retrieve the EventNumber variable in the main workflow. The reason for that is the following one: our analysis use the usual 50:50 event splitting between even/odd events (i.e. train on even events, then apply to odd ones and vice versa), therefore that variable is a must have to make the correct association. The critical point is that it is retrieved and casted as a simple integer from the input container, while it should be a ULong64_t as defined in the proper container. In case of samples with limited statistics, it isn't really a problem, but for larger samples, that can be problematic if the analysis rely on that variable, because it is stored as an integer in the output ntuple.

Please, let me know if you have comments or questions related!