fix: disable thinking mode in Qwen35VL single-episode fallback path

The single-episode `segment_skills` method was missing
`enable_thinking=False` in `apply_chat_template`, causing the model to
output reasoning traces instead of JSON when the batch path fails and
falls back to per-episode processing.

Made-with: Cursor

src/lerobot/data_processing/data_annotations/vlm_annotations.py

@@ -495,7 +495,9 @@ class Qwen35VL(BaseVLM):
             },
         ]
 
-        text = self.processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
+        text = self.processor.apply_chat_template(
+            messages, tokenize=False, add_generation_prompt=True, enable_thinking=False
+        )
         image_inputs, video_inputs = self.process_vision_info(messages)
         inputs = self.processor(
             text=[text],