Add method stir.ToPriorWithParabolicSurrogate() to python.

[robbietuk]

Input is a GeneralisedPrior. This will give access to methods such as parabolic_surrogate_curvature and add_multiplication_with_approximate_Hessian from the objective function object.

This would be a temporary fix for part of UCL#707

This doesnt expose non-parabolic surrogate type function methods (e.g. get_gamma() and get_epsilon() in the RDP)

[robbietuk]

I have tested this method on the QP. The code was

para=image.get_empty_copy()
stir.ToPriorWithParabolicSurrogate(self.recon.get_objective_function().get_prior()).parabolic_surrogate_curvature(para, image)
from show_estimate import *
show_estimate(image, title="image")
show_estimate(para, title="parabolic_surrogate_curvature")
stir.ToPriorWithParabolicSurrogate(self.recon.get_objective_function().get_prior()).add_multiplication_with_approximate_Hessian(para, image)
show_estimate(para, title="add_multiplication_with_approximate_Hessian")

which assumes the image is a STIR container, and self.recon is an OSMAPOSL object.

The primary STIR method is stir.ToPriorWithParabolicSurrogate() which accepts a GeneralisedPrior object. This then gives access to parabolic_surrogate_curvature() and add_multiplication_with_approximate_Hessian() methods.