On the Analysis of Computational Delays in Reinforcement Learning-based Rate Adaptation Algorithms