Use cli for all operations instead of the API

We have to re-implement logic (like updating a workspace before starting it) and it would be better to just call out to the cli since it already accounts for all that.

It is a lot of surface area to cover (every plugin) any time a feature is added.

If the command errors, show whatever is in stderr as the error message.

An alternative would be to handle these things at the API layer, and maybe there is a case to be made for that.

首先定位目前重新實作 workspace 更新和其他 API 行為的外掛程式操作，然後檢查如何呼叫 CLI。定義涵蓋所有外掛程式的範圍，並驗證 CLI 失敗時會將 stderr 暴露為錯誤訊息；該 issue 尚未解決 API 層的替代方案。

由索引模型根據 Issue 內容生成。