Add the mean yield calculation in post-fit sampling

[anigamova]

With this setup I can reproduce the post-fit uncertainties calculated with FitDiagnostics. Thought about doing the same but in one pass, i.e. calculate sum(y^2) and define err as mean(y^2) - (mean(y))^2, but it looks like especially with large numbers the calculation with two loops is more accurate.

[anigamova]

@kcormi you are completely right, I don't think that it makes much of a difference in speed, but we might use yield distributions for further studies anyway, but I don't think that we should to save in the output files.

[anigamova]

Thanks for the feedback @ajgilbert, everything is implemented now.