Finite-Time Analysis of Asynchronous Q-Learning with Discrete-Time Switching System Models