Spectral method solver (exploratory)

[dschwen]

Now that we build Magpie with FFTW I'd like to do some exploratory work to see how feasible a spectral solver (for phase field equations for example) would be. This would require structured orthogonal grid meshes and at first will be serial only. FFTW3 supports MPI parallelization, and I will work on this later. It will require a custom partitioner to partition the mesh according to the FFTW domain decomposition (which is a sliced domain in one direction - quite simply).