merge from linearAlg-branch. my parallelization needed a slightly more general linear solver interface and uses the pvd-naming scheme from the petsc-parallelization