Learning Agents of Bounded Rationality: Rewards Based on Fair Equilibria