Improve Ikeda-Carpenter fitting

[Anders Markvardsen]

This includes adding constraints to mixing parameter Eta to place it within the bounds of zero and one.

[Anders Markvardsen]

(In [4399]) Include constraint to mixing parameter Eta to place it within the bounds of zero and one. Refs #1233

[Anders Markvardsen]

(In [4436]) Mathematica fix so that constraints not needed for IC Eta parameter. Refs #1233.

[Anders Markvardsen]

Previous fix improved stability of IC.

However, when translating parameters values from IRF Fullprof file not sure if this is done correctly.

I have so far assumed these are related to SigmaSquared and Gamma in the following way:

SigmaSquared = Sig-0 + d^{2 * Sig-1 + d}4 * Sig-2 Gamma = Gam-0 + d * Gam-1 + d^{2 * Gam-2}

Is this correct? Have asked Laurent about this.

Also can a value for Eta be estimated from the IRF?

[Anders Markvardsen]

(In [4752]) Problem with IC function returning NaN when x-axis interval too large. Fix this problem by using gsl_sf_log_erfc instead of gsl_sf_erfc. Refs #1233.

[Anders Markvardsen]

(In [4790]) Added numerical derivative method (similar to that in UserFunction) which can optionally be called from a functionDeriv() function implementation. Added such an implementation for IC. Refs #1233.

[Anders Markvardsen]

(In [4791]) Fix tests which were broke with previous commit to this ticket. Refs #1233.

[Anders Markvardsen]

(In [4806]) Improved speed at which numerical derivs calculated. Test with double peak Ikeda-Carpender gave good result. Refs #1233.

[Anders Markvardsen]

(In [4807]) Fix test. Refs #1233.

[Stuart Campbell]

This ticket has been transferred to github issue ​2080