Updates to configuration/categories/ML

[mafrahm]

Changes

• small configuration changes
  • improved dataset configuration using if_era function
  • colors, groups, ...
• Categories
  • in DL, we now create three types of categories based on mll (below/above/within Z mass peak)
• Machine Learning
  • We can now reuse MLPreTraining for models with different process groups
  • fixed some issues where shuffling was missing after merging multiple datasets into one process --> now each sub-dataset should be equally distributed over the train/val/test sets
  • allow using full event weight in training
  • adds some new models with HH vbf and single H included
  • some small but important bugfixes
  • CustomMergeMLEvents (not yet used, but should be faster as far as I can see)