This PR makes a few related changes:
1) Allow move semantics for GPUArray so we can return them from a creator function.
2) Create PSF.copy_to_gpu() function that allocates a GPU array and copies the data.
3) Change the remaining GPU memory functions to use GPUArray so that they pick up the error checking and validation.
4) Clean up cuda_errors.h which is no longer used.
This PR makes a few related changes: 1) Allow move semantics for
GPUArrayso we can return them from a creator function. 2) CreatePSF.copy_to_gpu()function that allocates a GPU array and copies the data. 3) Change the remaining GPU memory functions to useGPUArrayso that they pick up the error checking and validation. 4) Clean upcuda_errors.hwhich is no longer used.