Avoiding the boson sign problem at finite chemical potential