fix(annotate): don't crash pipeline on persistent JSON parse failure

Some prompts/models occasionally return pure prose with no JSON object
even on retry. Returning None (and logging a preview) lets the pipeline
skip that one VLM call cleanly instead of aborting the whole episode.
The modules already check for None / non-dict results and degrade
gracefully (no row emitted from that call).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

src/lerobot/annotations/steerable_pipeline/vlm_client.py

@@ -164,7 +164,17 @@ class _GenericTextClient:
                 },
             ]
             retry_text = self.generate_text([retry], max_tok, temp)[0]
-            out.append(_strip_to_json(retry_text))
+            try:
+                out.append(_strip_to_json(retry_text))
+            except (ValueError, json.JSONDecodeError):
+                # After retry: log preview and return None instead of crashing
+                # the whole pipeline. Modules treat None as "skip".
+                preview = retry_text.strip().replace("\n", " ")[:200]
+                print(
+                    f"[vlm] WARNING: failed to parse JSON after retry; preview: {preview!r}",
+                    flush=True,
+                )
+                out.append(None)
         return out