uniform pdfs for rateParam & flatParam nuisance parameters

[nucleosynthesis]

In most use cases, there is no need to include a pdf for rateParam / flatParam nuisances (in all fully frequentist methods, it makes no difference), however having the ability to include them could be useful in some cases eg

• Bayesian methods in which the nuisance parameter (not just the POI) has a uniform prior
• Cousins-Highland based toy generation where one or more nuisance parameter is assumed to have a flat pdf

It should be as simple as creating a user flag (even rtd) that tells combine to create a uniform pdf and include it into the NuisancePdf (along with some dummy global observable that effectively does nothing), but would also require some basic testing to make sure it doesn't break other things.

[nucleosynthesis]

More thoughts on this ... it should be possible to use the fact that these nuisances if declared as such in the card, will have the attribute "flatParam". This could be used to check which additional Prior terms should be constructed during the text2workspace stages.

[anigamova]

It is unclear to me why do we need a global observable or what to do with it in this case. Obviously RooUniform does not have any parameters so we can not introduce a global observable to the function as it is done in RooGaussian for example.

[kcormi]

Addressed by #834