Fix cancel generation feature by removing GeneratorExit handler

- Removed GeneratorExit exception handler from UI layer that was preventing cancellation
- Allow GeneratorExit to propagate to backend generator where cancel_event checks work
- Cancel button now properly stops text generation mid-stream
- UI state resets correctly after cancellation

app.py

@@ -665,15 +665,6 @@ with gr.Blocks(title="LLM Inference with ZeroGPU") as demo:
                     chat: response_chunk[0],
                     dbg: response_chunk[1],
                 }
-        except GeneratorExit:
-            # Handle GeneratorExit by resetting UI and returning cleanly
-            print("Generation cancelled by user.")
-            yield {
-                txt: gr.update(interactive=True),
-                submit_btn: gr.update(interactive=True),
-                cancel_btn: gr.update(visible=False),
-            }
-            return  # Don't re-raise, just return cleanly
         except Exception as e:
             print(f"An error occurred during generation: {e}")
             # If an error happens, add it to the chat history to inform the user.