Adding optimisation of diphoton mass resolution categorisation

[JaLuka98]

The categorisation of events is very important to enhance the sensitivity. In the case of the fiducial/differential analysis (c.f. e.g. CMS-HIG-19-016), we categorise in the diphoton invariant mass resolution.

The optimisation algorithm is given in the thesis by Thomas, chapter 6.5.2 (whole of chap 6.5 is very helpful). We should implement it like this!

A further step has to be included as well: The decorrelation of sigma_m_over_m against m_yy itself to avoid mass sculpting effects. This is usually done using the Diphoton Sherpa samples. They are currently still being produced but first files are coming in, see the DAS link. For the optimisation itself, also G+Jet samples appear to be used.

Additions should go to this branch: dev_higgsdnafinalfit

NB: Related to HiggsDNA-MR67