Optimal software release policy with a learning factor for imperfect debugging