Add option to use either cudaMalloc or cudaMallocManaged

[WilliamTambellini]

Hi @ggerganov
Would you mind if I add an option to possibly use
cudaMallocManaged() instead of cudaMalloc() in
https://github.com/ggerganov/ggml/blob/master/src/ggml-cuda.cu
default being of course cudaMalloc()
?
Best
W.
refs:
https://developer.nvidia.com/blog/unified-memory-cuda-beginners/
https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__MEMORY.html#group__CUDART__MEMORY_1gd228014f19cc0975ebe3e0dd2af6dd1b