Computation of the Asymptotic Bias and Variance for Simulation of Markov Reward Models