Implementation of theory-motivated agnostic bands

[emanca]

The bands can be produced using the script makePdfUncPlot.py and then are used by default by setupCombine to build the shapes. It's possible to increase the size of the band using the new option theoryAgnosticBandSize which defaults to 1. The size of the OOA is hardcoded to 1 and the variations on sigmaUL are always 50%, even if the band is increased. I also corrected a problem in the implementation that prevented the fit to run correctly when --hdf5 option is used. Everything has been tested locally and it works and the output makes sense. We wish to merge this ASAP (at most tomorrow morning) to run the bias tests with @jeyserma together with the traditional analysis, therefore @cippy and @davidwalter2 please tell me what you think.

[cippy]

From a first look it would seem fine, and it shouldn't modify other things run in CI, but in any case you can surely start running all tests you need even before the PR is merged, since the main code with this PR will simply be your local code.

[emanca]

From a first look it would seem fine, and it shouldn't modify other things run in CI, but in any case you can surely start running all tests you need even before the PR is merged, since the main code with this PR will simply be your local code.

Of course, but @jeyserma is currently running the bias tests on both setups

[bendavid]

Is this ready to merge?

[bendavid]

From the ci only the theory agnostic changes as expected (with the norm impacts getting significantly smaller, but probably it's hard to draw any conclusions from the limited stat run in the ci)

[bendavid]

ci results still look consistent