User Input AiAction responses from the websocket break AI outputs

In this slack conversation @rdickenson detected a console error relating to an ExplainVulnerability response being returned with a null errors key, resulting in a parsing failure, however the interface was working as expected from what we could see.

However we noted multiple responses with the same request id being returned for the same AiAction request, as can be seen in this screenshot:

It would appear this is due to Trigger subscription message for user input (!122647 - merged) • Jan Provaznik • 16.3 in which the user's interactions now result in websocket outputs, which other AI features such as Explain This Vulnerability and Explain Code did not expect, resulting in the feature name being returned as the responseBody and displayed as if it was the AI response.