Add test priors 2

[robbietuk]

This PR is to replace and expand upon that of #519.

Addresses #153

• add Logcosh Log cosh prior #716
• consider adding Hessian test
• consider adding surrogate test

[robbietuk]

Travis will fail on PLS because of edge artefacts. Here is a difference image. PLS uses a uniform 1's anatomical image and the same randomised densitity (between 0 and 1 voxel values) as QP and RDP

Codacy is still not happy for the same reasons as mentioned #519 (comment)

[robbietuk]

I believe PLS is failing at the (max) edge because of continue statements in the directional gradient computation

STIR/src/recon_buildblock/PLSPrior.cxx

Lines 318 to 333 in 7b4c274

	if(direction==0){
	if(z+1>max_z)
	continue;
	image_gradient_elem[z][y][x]=image[z+1][y][x]- image[z][y][x];
	}
	if(direction==1){
	if(y+1>max_y)
	continue;
	image_gradient_elem[z][y][x]=image[z][y+1][x]- image[z][y][x];
	// std::cout<<"grady ="<<image[z][y+1][x]- image[z][y][x]<<std::endl;
	}
	if(direction==2){
	if(x+1>max_x )
	continue;
	image_gradient_elem[z][y][x]=image[z][y][x+1]- image[z][y][x];

The solution to this is not obvious.

The failure at the other edge (top and left) is harder to reason but I have seen similar issues with boundary conditions with RDP and QP. I have not had a chance to look through the PLS code.

[KrisThielemans]

@robbietuk, what about disabling the PLS test for now, adding logcosh and merging this PR?

[robbietuk]

Yes, I think we also expand the testing for RDP and logcosh priors to have more than one evaluation with different gamma(RDP)/scalar(logcosh) values. We didn't spot the logcosh error regarding the value compuation when the input is large for a couple of weeks.

Ill get on this shortly, did you want it for release4.1?

[robbietuk]

test_priors passes on my machine (waiting on Travis to confirm). This PR needs to be followed up with a further issue relating to additional tests for other prior methods, but these should wait until the mess with the priors code has been cleaned up (#694). The tests should also be expanded to vary paramerter for other methods.