feat(example-iot-telemetry): implement polyglot ObjectUri pipeline and Gemini AI anomaly synthesis

[crapougnax]

Part of parent story #41.

🎯 Goal

Implement polyglot ObjectUri routing across storage engines and cognitive incident diagnosis via Google Gemini AI.

📋 Specification

1. Polyglot ObjectUri resolution:
   • Device model: postgres:devices/dev-sensor-42.
   • Message in transit: amqp:sensor-frames/msg-881.
   • Cold storage raw payload: s3:telemetry-cold/2026-09/dev-sensor-42.json.
2. Cold Storage Archiving:
   • Export raw batches to S3 / MinIO via @quatrain/storage-s3.
3. AI Cognitive Anomaly Synthesis:
   • Integrate @quatrain/ai-gemini to detect critical threshold anomalies (e.g. abrupt soil moisture drop, battery sag).
   • Use generateStructured() with strict JSON schema to produce actionable diagnostic cards (anomaly kind, probable root cause, severity, recommended maintenance action).