Suboptimal policies for stochastic N-stage optimization problems: Accuracy analysis and a case study from optimal consumption