Function evaluate_task_progress 

pub async fn evaluate_task_progress(
    ctx: &TaskLoopContext,
    session: &Session,
    llm: Arc<dyn LLMProvider>,
    event_tx: &Sender<AgentEvent>,
    session_id: &str,
    model: &str,
    reasoning_effort: Option<ReasoningEffort>,
) -> Result<TaskEvaluationResult, AgentError>

执行 TaskList 评估