GCP-08: Cloud Workflows for River of Intelligence — Declarative Pipelines

Replace river_scheduler.py procedural orchestration with Google Cloud Workflows for declarative, observable, retryable pipeline execution. Current pain: river_scheduler.py crashes mid-pipeline with no auto-resume. Gemini API 60-120s calls make it fragile. Architecture: 1. Workflow: river-of-intelligence with steps: ingest -> classify -> parallel(enrich, learn) -> pulse -> summary 2. Triggers: Cloud Scheduler daily at 02:00 UTC, manual, event-driven on new products 3. Per-step retry with backoff, visual execution trace in Cloud Console 4. River worker refactored to HTTP handlers per stage Deliverables: - river_of_intelligence.yaml (Workflows definition) - Refactored river_worker.py (HTTP handlers per stage) - Cloud Scheduler daily trigger - Migration guide for parallel run with existing scheduler Target: c8-infra/workflows/river_of_intelligence.yaml + c8_platform/river_worker.py