Optimal scheduling policies for diagnosing k-out-of-n systems with imperfect tests