Discrete time Markov reward processes as stochastic annuities