Add eval harness (#62 P4), status frames (#96), and agentic runs (#63).

Ship the Phase 4 accuracy eval suite with a manual Gitea workflow, emit
versioned WS status frames during grounded chat, and introduce opt-in
agent infrastructure (Redis/Celery, AgentRun/Step, tools, orchestrator)
gated by ALLOW_AGENTIC_TASKS so default chat behaviour stays unchanged.

llm_be/monetization/models.py

@@ -87,6 +87,9 @@ class SubscriptionPlan(TimeInfoBase):
             return self.allows_image_generation
         if feature in ("rag", "document_rag"):
             return self.allows_rag
+        # Agentic multi-step runs (#63): same tier as RAG until a dedicated flag.
+        if feature in ("agentic", "agentic_tasks"):
+            return self.allows_rag
         return False