Add mimizer interface to be used by Fit

[Anders Markvardsen]

Add IFuncMinimizer interface to be used by Fit

This interface will have methods which are independent of the minimizer used. Methods for this interface include: iterate() and hasConverged().

Also add concrete implementations of this interface for: Levenberg Marquardt Simplex Fletcher-Reeves/Polak-Ribiere conjugate gradient BFGS

[Anders Markvardsen]

(In [3650]) Renamed method in Fit nParams to nActive and removed unnecessary copying operation in preparation to use IFuncMinimizer in Fit. Refs #1080.

[Anders Markvardsen]

(In [3651]) Test fail on Linux temporarily committed out. Refs #1080.

[Anders Markvardsen]

(In [3657]) Integration of LevenbergMarquardtMinimizer into Fit. Refs #1080.

[Anders Markvardsen]

(In [3661]) Adding SimplexMinimizer. Refs #1080.

[Anders Markvardsen]

(In [3673]) Adding derivative of cost function to Fit and Fletcher-Reeves conjugate gradient minimizer. Set property "Minimizer"="Conjugate gradient (Fletcher-Reeves imp.)" to use. Refs #1080.

[Anders Markvardsen]

(In [3674]) Increase tolerance back up for gaussian simplex test to get test accepted on both build server and local machine. Refs #1080.

[Anders Markvardsen]

(In [3681]) Adding Polak-Ribiere flavour conjugate gradient and Broyden-Fletcher-Goldfarb-Shanno (BFGS) minimizer. Set property "Minimizer"="Conjugate gradient (Polak-Ribiere imp.)" or "Minimizer"="BFGS" to use. Refs #1080.

[Anders Markvardsen]

(In [3694]) Calculate covariance matrix output using new minimizer interface. For this is only done when methodUsed="Levenberg-Marquardt". Also still need code for the case when the specified parameters by the user are not the same as the parameter fitted on. Refs #1080.

[Anders Markvardsen]

(In [3696]) Fit tests broken in previous commit to this ticket. Refs #1080.

[Roman Tolchenov]

(In [3698]) Added code calculating the standard deviations of transformed parameters. re #1080

[Anders Markvardsen]

(In [3703]) BackToBackExponential test back to working again after adding GSL fix to LevenbergMarquardtMinimizer. Refs #1080.

[Anders Markvardsen]

(In [3710]) Code add for calculation covariance matrix for conjugate gradient minimizers. Refs #1080.

[Anders Markvardsen]

(In [3711]) Roman check that covariance matrix calculation for conjugate gradient minimizers is OK, hence here adds the same kind of code for the BFGS minimizer. Refs #1080.

[Anders Markvardsen]

(In [3712]) Forgot to commit this Fit.cpp change with previous commit to this ticket. Refs #1080.

[Stuart Campbell]

This ticket has been transferred to github issue ​1928