Constrained Reinforcement Learning with Average Reward Objective: Model-Based and Model-Free Algorithms