The simplex and policy-iteration methods are strongly polynomial for the Markov decision process with a constant discount