Feature/add cuda

[corentin-dev]

This is a pull request to had support for CUFFT and CUDA in PencilFFTs.

CUFFT does not accept some FFTW so I had to make a specific FFTW call to plans.

I also changed a bit some of PencilArrays in order to be able to work with CuArrays.

Many changes on how internal vectors are allocated in order to be able to allocate CuArrays.