gpuarray supported forward and backprojectors

I have code from a contributor of TIGRE that allows, for users with the distributed computing toolbox, and for CT images that fit inside the GPU together with the algorithm, to avoid CPU-GPU transfers and to everything in the GPU.

I don't have the physical time to add them to TIGRE, but if someone wants to, happy to share it and give a hand on the design on how to integrate it seamlessly into current TIGRE.