todo mpi

MPI

The code is currently only thread-parallel. We need to read in a partitioned grid and implement communications in the flux (and flux-jacobian) computation, gradient computation and solution reconstruction routines.