fix(structured): harden structured output for local servers and thinking models

- Local servers (LM Studio, vLLM) reject the object-form tool_choice langchain
  sends for function calling. The generic openai_compatible provider now binds
  the schema as a tool without forcing tool_choice.
- A structured call can return no parsed result (a thinking model answering in
  plain text); fall back to free text with a clear reason instead of an opaque
  render error.

tradingagents/agents/utils/structured.py

@@ -63,6 +63,11 @@ def invoke_structured_or_freetext(
     if structured_llm is not None:
         try:
             result = structured_llm.invoke(prompt)
+            if result is None:
+                # A thinking model can answer in plain text instead of calling
+                # the tool, leaving the parser with nothing to return. Treat it
+                # as a structured miss and fall back, with a clear reason.
+                raise ValueError("structured output returned no parsed result")
             return render(result)
         except Exception as exc:
             logger.warning(