CUDA full n-dimensional support

[erikzenker]

Currently buffer copies are only defined for up to three dimensions. But coming unit tests will use at least four dimensional buffers. Therefore, we need to add support for n-dimensional copy operations.

[BenjaminW3]

#290 implements full n-dimensional support for all missing CPU operations (set and copy). N-Dimensional CPU execution support was already there.

Still missing is the support for more then 3 dimensions for CUDA kernel execution (ExecGpuCudaRt::Enqueue, AccGpuCudaRt::getAccDevProps, IdxBtCudaBuiltIn, IdxGbCudaBuiltIn) as well as all cuda operations (BufCudaRt::copy, BufCudaRt::set, BufCudaRt::alloc) and maybe more.