A new stopping criterion for iterative solvers for control optimal problems