Improve fitting

[peterfpeterson]

Current peak fitting does not give terribly good results in the general case. Certain things can be done to make it better:

1. Improve initial guess of peak location through either starting with the position of the maximum, or by calculating the first moment of the function
2. Modify mantid's FitPeaks to better set initial guesses for pseudo-Voigt. The issue is that it sets things using effective parameters and pseudo-Voigt needs the mixing parameter set first.

[JeanBilheux]

Values and error return by get_effective_params() are identical, except for the Height.

Values

Errors

[peterfpeterson]

The issue Jean is pointing out is specifically for the pseudo-Voigt. Gaussian is fine.

[peterfpeterson]

A separate issue has been created for the pseudo-Voigt uncertainties #303

[peterfpeterson]

mantid PR27809 fixes the last bits of this issue.