Implement GPU-based Morlet filtering

[mivade]

After cleaning up the C++ Morlet wavelet code, it would be nice to optionally have a GPU implementation and benchmark it against other methods. I would suggest using Boost.Compute which is backed by OpenCL. An alternative is ArrayFire which can use CPU, OpenCL, or CUDA depending on what is available. The major downside here is it's an additional library to link against (cf. Boost.Compute which is header-only apart from the OpenCL requirements).