Allow LLMs to read correct success or error from scripts Stop action

[super-qua]

Describe the feature

LLMs in Assist do not receive the correct information if a tool call was successful in case of a script.

Adding a Stop action to the script with

stop: "Error: error"
error: true

does not return any error to the llm in the tool call result. Instead, in https://github.com/home-assistant/core/blob/a265ecfade21520742f6bc2226db654e3c571f00/homeassistant/helpers/llm.py#L1005, the success field is hard-coded to True, and the stop and error fields from the Stop action are not returned.

Therefore the LLM Assistant is unable to trying to correct the tool call, or to give a correct response to the user in case of an error.

Use cases

LLM Assistants running any user script as a tool.

Anything else?

No response