Keep the Cost Down: A Review on Methods to Optimize LLM' s KV-Cache Consumption