Per request from the RHDH LS team, we need to create a message level inline error state for failed instances.
example: when a user waits for the AI message streaming to complete but for some reason the streaming is stopped. In this case we may want to surface the error happened only on the last message from the bot/AI.

