A data engineer is designing a streaming ingestion pipeline using Auto Loader. The requirement is that the pipeline should never fail on schema changes but must capture any new columns that arrive in the data for later inspection. Which of the following schema evolution modes should the engineer use to meet this requirement?