Fix: Pass user query to LLM context in Fast Lane and Deep Dive stages

Upgrade to Hybrid Architecture v3 (Fast Lane + Deep Dive)

Remove conversation history from LLM context to reduce tokens

Add compact prompts and context builder - reduces token usage by ~50%

Fix: pass context to _generate_response for persona/zone/trend/weather

Full Developer Spec implementation: Priority-based Multi-Stage Reasoning

Implement multi-stage reasoning architecture (Claim→Validate→Contradict→Confirm)