Fixing MergeCategorization

Fixing MergeCategorization:

• When merging nanoaod files, if the branches differ between files (e.g. in data, different HLT paths available), the events were skipped. This is solved by using haddnano.py instead of hadd.
• However, using haddnano.py seems a bit slower, so haddnano.py is used by default for data and hadd stays as default for MC, where in general this issue should not appear.
• In any case, if the user thinks hadd or haddnano.py can be used for the dataset he/she wants to merge, the parameters --force-hadd and --force-haddnano can be included in the MergeCategorization command.

Apart from this, included a new CMT_STORE_EOS_MERGECATEGORIZATION environment variable to store the output from this task. This implies some changes in the setup file, as shown in this MR.