Algorithms and Bounds for Rollout Sampling Approximate Policy Iteration