Add multi-pattern orchestrator with plan-then-execute and supervisor (#739)

Introduce an agent orchestrator service that supports three
execution patterns (ReAct, plan-then-execute, supervisor) with
LLM-based meta-routing to select the appropriate pattern and task
type per request. Update the agent schema to support
orchestration fields (correlation, sub-agents, plan steps) and
remove legacy response fields (answer, thought, observation).

trustgraph-base/trustgraph/base/agent_client.py

@@ -57,8 +57,7 @@ class AgentClient(RequestResponse):
         await self.request(
             AgentRequest(
                 question = question,
-                plan = plan,
-                state = state,
+                state = state or "",
                 history = history,
             ),
             recipient=recipient,

trustgraph-base/trustgraph/base/agent_service.py

@@ -90,9 +90,6 @@ class AgentService(FlowProcessor):
                         type = "agent-error",
                         message = str(e),
                     ),
-                    thought = None,
-                    observation = None,
-                    answer = None,
                     end_of_message = True,
                     end_of_dialog = True,
                 ),