A data engineer is exploring schema evolution modes in Databricks Auto Loader. Which mode should they select to ensure the stream continues without failure upon encountering a new column?