Average-Reward Off-Policy Policy Evaluation with Function Approximation