NVCC can compile Boost::Python modules

In this [discussion of using thrust with pycuda](http://pycuda.2962900.n2.nabble.com/PyCUDA-Compiling-thrust-code-in-pyCUDA-td7571922.html) a few years ago, it was stated that using multiple modules while creating both a Boost::Python module and cuda code was necessary because nvcc couldn't handle boost.

I can now compile a simple Boost::Python module together with pycuda code, as seen in [this gist](https://gist.github.com/kazimuth/f9952810d8117f782be71279c0f16f6e).

It might be possible to simplify [cuda.py](https://github.com/inducer/codepy/blob/master/codepy/cuda.py) now?