Add fixed chemical potential as an alternative constraint

We currently fix particle numbers, which is fine for closed systems. Fixed chemical potential is useful for open systems and then we could also deal with non-hermitian Hamiltonians (e.g. systems with leads).