Leveraging Reinforcement Learning in Red Teaming for Advanced Ransomware Attack Simulations