Early return for num_shards==0 in the Beam pipeline.

python/mlcroissant/mlcroissant/_src/operation_graph/execute.py

@@ -195,6 +195,9 @@ def execute_operations_in_beam(
         enumerate(files)
     )
     num_shards = len(files)
+    if not num_shards:
+        logging.warning(f"Empty {record_set=} for filters={json.dumps(filters)}")
+        return pipeline
 
     # We don't know in advance the number of records per shards. So we just allocate the
     # maximum number which is `sys.maxsize // num_shards`. Taking the practical case of