fix: initialize agent properly in API mode

Fixed critical bug where ModelDayExecutor._initialize_agent() created
a BaseAgent but never called agent.initialize(), leaving self.model=None.
This caused 'NoneType' object has no attribute 'bind' error when
run_trading_session() tried to create the langchain agent.

Changes:
- Made _initialize_agent() async
- Added await agent.initialize() call
- Updated call site to await async function

Now properly initializes MCP client, tools, and AI model before
executing trading sessions, matching the working CLI mode pattern.

api/model_day_executor.py

@@ -126,7 +126,7 @@ class ModelDayExecutor:
             os.environ["RUNTIME_ENV_PATH"] = self.runtime_config_path
 
             # Initialize agent
-            agent = self._initialize_agent()
+            agent = await self._initialize_agent()
 
             # Run trading session
             logger.info(f"Running trading session for {self.model_sig} on {self.date}")
@@ -209,7 +209,7 @@ class ModelDayExecutor:
         """Execute model-day simulation (sync entry point)."""
         return self.execute_sync()
 
-    def _initialize_agent(self):
+    async def _initialize_agent(self):
         """
         Initialize trading agent with config.
 
@@ -259,6 +259,9 @@ class ModelDayExecutor:
         # - Database initialization is handled by JobManager
         # - File-based position tracking is only for standalone/CLI mode
 
+        # Initialize MCP client and AI model
+        await agent.initialize()
+
         return agent
 
     def _create_trading_session(self, cursor) -> int: