* Add GPU context initialization (cuda_enter) and finalization (cuda_exit)
* Add proper synchronization on CUDA stream (cuda_wait) before and after performing operations
* Reorder C variables at the start of the function to improve readability
* Add ext_cuda initialization to utilize the synchronization and context functions in the GPU
Signed-off-by: João Victor Tozatti Risso <joaovictor.risso@gmail.com>