KV cache checkpointing is a powerful optimization technique for agentic AI sessions because it directly addresses the unique computational and memory challenges