Update sample list for new production

• add 2016, 2017 (new 2018) samples
• change script to insert file lists:
  • add "eras" entry if a same sample exist for several years (to avoid repeating the cross sections etc.)
  • samples can be recognized using regexes (to handle differences across years, extensions etc.)
  • sum of weights for template normalization is taken from sample json (for now) because of bug in the preprocessing

The job splitting will certainly have to be tuned once we run bamboo...