Distropy Rust server hits 60k t/s prefill

The 60,000 tokens per second claim is a specialized benchmark for KV prefix caching, but it highlights a critical bottleneck in local LLM deployment. By making large system prompts and tool schemas virtually free after the first request, Distropy's Rust-based implementation offers a significant performance and memory safety advantage over Python-centric alternatives like vLLM. It targets the context bloat in IDE extensions where prefill delays can reach ten seconds, enabling near-instantaneous multi-turn conversations and rapid agentic tool-calling on consumer hardware like the RTX 4070.