RL in Markov Games with Independent Function Approximation: Improved Sample Complexity Bound under the Local Access Model.