additional topology-aware handling for Pstream

• extend base MPI routines to include a two-stage handling from inter-node to local-node
• adjust manual algorithms to include two-stage inter-node/local-node handling