fix: sanitize messages and ensure 'content' for strict LLM providers

- Strip non-standard keys like 'reasoning_content' before sending to LLM
- Always include 'content' key in assistant messages (required by StepFun)
- Add _sanitize_messages to LiteLLMProvider to prevent 400 BadRequest errors

nanobot/agent/context.py

@@ -227,9 +227,9 @@ To recall past events, grep {workspace_path}/memory/HISTORY.md"""
         """
         msg: dict[str, Any] = {"role": "assistant"}
 
-        # Omit empty content — some backends reject empty text blocks
-        if content:
-            msg["content"] = content
+        # Always include content — some providers (e.g. StepFun) reject
+        # assistant messages that omit the key entirely.
+        msg["content"] = content
 
         if tool_calls:
             msg["tool_calls"] = tool_calls