Investigate the use of PyOpenCL for faster parallelization

[monocongo]

I've recently had good luck with OpenCL in the context of computer vision applications (using UMat objects with OpenCV), and thought it'd be interesting to cook OpenCL into this codebase to see how much better performance we can get from it as a result.

Some references to help get started: PyOpenCL documentation Parallel Programming with (Py)OpenCL for Fun and Profit