Optimality of Reinforcement Learning Algorithms with Linear Function Approximation