Finite-Time Analysis of Asynchronous Q-learning under Diminishing Step-Size from Control-Theoretic View