Code MPI-fähig machen

TODO:

• (!91 (merged)) Matrix-free host
• [ ] Matrix-base host
• (!95 (merged)) Device
• (!95 (merged)) Device-v2
• (!97 (merged)) Fix timings code to deal with MPI
• (jfahl_01/pps-benchmarks!14) Benchmark (non-mpi only)
• [ ] What about benchmarking MPI? (Hast to wait until we have cluster access)