Dynamic/Automatic blinding in plotter

Currently blinding usually relies on analyzers introducing some fixed blinding cuts to avoid looking at data in signal sensitive regions. In principle one can define some simple criteria such as s/b and s/sqrt(s+b) with some conservative (=low) thresholds to determine which bins in a histogram may be sensitive and thus should be blinded.

In analyses throwing essentially everything into a DNN it may be difficult to judge the modeling of the individual features if some of them may already provide "visible" sensitivity to a potential signal. Even more so, control regions may only be defined via the same DNN (multiclassifier or a separate DNN using similar/overlapping sets of features). In these cases it would also be difficult to check for potential modeling issues without manual fine-tuning of blinding criteria for every feature. Hence, the plotter feature described above can be very useful for such analyses.

TODO: add options to the plotters to automatically blind data in bins where s/b and/or s/sqrt(s+b) exceed a configured threshold (or where b==0 as a safety measure in case signals may be in a very far-out phase space and to avoid division-by-zero issues in s/B for example)