A data engineer is exploring schema evolution modes in Databricks Auto Loader. Which mode should they choose to ensure the stream continues without failing, while also ignoring any new columns that may appear?