README: explicit model labels — 1.7B study vs 0.6B study index

Add Qwen3-TTS 0.6B load-test report: 1.7B comparison + TRT-vs-eager vocoder A/B (RTX PRO 6000)

Add exact TRT vs native vocoder FastAPI and Silero A/B metrics

Add TRT vocoder vs raw vLLM benchmark report

Upload VOCODER_RUNTIME_BENCH.md with huggingface_hub

initial commit