[WIP] Extra options for nuisance parameters and backgrounds treatment

[Revile77]

This PR is large, so apologies for convoluting so many changes in a single PR ... I realized they depend on each other too much hence why. But we can of course try to iterate on them one by one.

Several features we found useful in a specific measurement are introduced here:

• By default ,we assume that the nuisance parameters for object-related systematic uncertainties have gaussian priors. For certain diagnostic cases, it is usefull to replace these with flat prior, the obj_syst_flatprior dictionary is added for this purpose, where every NP with a matching substring in the name is assigned a flat prior with configurable bounds. The design of this feature is pretty terrible from interface perspective, but unfortunately I find that anything more complex requires serious interface reworking, hence this preliminary support.
• Ability to fix NP to a certain value -- the motivation from my side is mostly the ability to estimate impact of NP on the result by fixing and varying it's value.
• Treatment of limited statistics available for the background estimate by introducing per-bin nuisance parameters with a Poisson constraint (somewhat similar to Beeston-Barlow approach). The implementation loosely follows that in the HistFactory documentation.