Top 10 KV Cache Compression Techniques for LLM Inference: Reducing Memory Overhead Across Eviction, Quantization, and Low-Rank Methods