Extend error calculation options

[woutergins]

As documented here, MINPACK can also search for asymmetrical errors by calculating the place where the function to be minimized reaches the value F_min + 1. For chisquare calculations, this is exactly what is needed, but loglikelihood calculations need to be checked to verify that this is indeed the correct approach. This way, we would give the option of calculating the errors in 3 ways, with increasing calculation time and accuracy.

method	Parabolic errors	Asymmetric errors	Distribution
chisquare	Yes	Yes	No
loglikelihood	No	Yes	Yes

[woutergins]

Reading the details of the implementation and interpretation of the error calculations of MINUIT here, it seems that a change of 0.5 of the loglikelihood gives the 1 sigma errors.

[woutergins]

This has now been implemented. The only missing function is to calculate the MINOS (asymmetric) errors, and save them in the parameters.