NPU LLM runner does not stop immediately when requested #64
Description
When the user tries to stop an in-progress LLM inference using NPU, the LLM runner continues executing in the background until completion.
This causes issues when the user immediately sends a new query, as the system may respond incorrectly.