Allows for CUDA registration for bittensor using python.

Cython
bittensor@feature/cuda_solver  
nvcc (cuda-11.6)
gcc (9.4.0)
sm_61 enabled CUDA GPU (10XX series or higher)

pip install -e ./bittensor_register_cuda

Testing uses unittest as there is an issue with pytest and Cython

python3 -m unittest test.py

• Speed-up transfer to host after finding a solution to POW
• Perhaps use events/streams

https://github.com/rmcgibbo/npcuda-example/
https://github.com/mochimodev/cuda-hashing-algos/