fix: increase OpenClaw bridge timeout + improve ask_openclaw error handling

feat: add OpenAI Vision API for camera analysis + fix OpenClaw bridge auth

Fix crash resilience and OpenClaw gateway connectivity

Update default OpenAI realtime model to gpt-realtime-1.5

Switch OpenClaw bridge from HTTP to WebSocket protocol

Add EMA smoothing to face tracking to reduce jitter

Fix missing Any import in main.py

Add local face tracking (YOLO) and vision processing (SmolVLM2)

Refactor to native OpenAI Realtime with OpenClaw identity

Fix speech cutoff and echo: track speaking state, wait for responses

Fix TTS: cancel OpenAI auto-response, route through OpenClaw

Add acknowledgment for long-running queries to prevent timeout

Fix modalities error: must use ['text', 'audio'] not ['audio']

Fix dual-response issue: disable OpenAI auto-response

Use x-openclaw-session-key header for proper session routing

Fix cross-channel conversation awareness with shared user ID

Fix conversation awareness across channels (WhatsApp, web, robot)

Add face tracking for robot to look at person speaking

Initial release of ClawBody