Chameleon: An AI System for Efficient Large Language Model Inference Using Adaptive Caching and Multi-Level Scheduling Techniques