Learning Infinite-Horizon Average-Reward Restless Multi-Action Bandits via Index Awareness