KV Cache Wars: How Compaction, Quantization, and a Rust Rewrite Are Redefining LLM Serving Economics