fix: use only field validator for scene_id conversion

Remove manual string conversion in parsing code and rely solely on the Pydantic field validator to handle integer to string conversion. This should fix CI test failures while maintaining support for integer scene_ids.

Author: PaulHax

align_browser/experiment_models.py

@@ -491,11 +491,6 @@ def add_experiment(
                 meta_info = item.input.full_state.get("meta_info", {})
                 if isinstance(meta_info, dict):
                     scene_id = meta_info.get("scene_id", f"scene_{source_index}")
-                    scene_id = (
-                        str(scene_id)
-                        if scene_id is not None
-                        else f"scene_{source_index}"
-                    )
 
             if scenario_id not in scenarios_dict:
                 scores_path = None