Refine avatar diagnostics and prune legacy assets

chore(ui): remove top bar placeholder buttons to focus on video diagnostics

feat(frontend): stage timeline, frame counter, black frame detection, connection path badge

feat: replace index.html with enterprise-grade UI

fix: improve webrtc server stability and diagnostics

feat: upgrade frontend to enterprise-grade premium design

WebRTC: add track-before-answer wait; remote autoplay/play(); relay-only fallback reconnect; mute remote video; minor client tweaks

Debug: add UI button to fetch /debug/models and log ONNX presence summary

UI: add Initialize AI Pipeline button; wire to POST /initialize with status updates

UI: fix Disconnect teardown; send reference immediately when connected; make avatar video fill container

feat(docker): switch to Docker Space GPU runtime; prod WebRTC (aiortc) flow; remove legacy WS; token auth; instrumentation p50/p95; requirements harden

Optimize for HuggingFace Spaces: simplified Gradio interface and reduced dependencies

feat: phase1-2 baseline echo scaffold