Support SuperLU_dist solver
Due to a demand of a PETSc maintainer for an issue I get with recent PETSc and Mumps in parallel.
It is anyway rather trivial to add a new solver supported by PETSc (even if it could be streamlined more - see #730 (closed))