Dynamic Operating System Scheduling Using Double DQN: A Reinforcement Learning Approach to Task Optimization