README: complete redesign with visual charts and viral formatting

- ASCII bar charts for KV quality comparison (quant.cpp vs llama.cpp)
- ASCII memory diagram showing the core problem/solution
- Scatter-style PPL comparison chart
- Collapsible FAQ sections (cleaner page)
- Collapsible architecture details (Gemma 4, performance breakdown)
- Centered hero with badge row (tests, score, WASM size, platforms)
- Side-by-side engine comparison table (vs llama.cpp/vLLM/MLX/ONNX)
- Model table with speed + compression columns
- Clear positioning: "Use llama.cpp for speed. Use vLLM for throughput.
  Use quant.cpp when you need more context in less memory."

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: unamedkr