Variable fit configurations

[eschanet]

I want to be able to do something like

simplify -f mu_SIG:1.0 inputLH.json > simplifiedLH.json

to run a fit with (in this case) the parameter mu_SIG is fixed (to 1.0 here).

This should be relatively simple to implement, just needs to propagate fixed parameters through fitter.fit() to pyhf like:

pyhf.infer.mle.fit(
        data,
        model,
        init_pars=init_pars,
        fixed_params=fixed_pars,
        return_uncertainties=True,
        return_result_obj=True,
    )

[eschanet]

Why do I want to have this? I want to be able to run e.g. a signal+bkg fit in order to construct the fitted bkg yields and uncertainties that are being plugged into the simplified LH.

This is interesting because a signal+bkg fit might actually have a significantly different behaviour than a bkg-only fit, meaning that the fitted background model plugged into the simplified LH might be different.

One thing to keep in mind for this specific example is that I will need to have a way of removing signal yields and uncertainties from total fitted yields. Removing nominal yields is trivial, but uncertainties is not because I would need to have the sample-wise uncertainties from pyhf.

[eschanet]

#7 merged so it can be used, however still need to solve above issue of removing the signal uncertainties in case user decides to run mu=1 fit and wants to remove the signal sample.