Consider a metastore Parquet table that
- doesn't have schema evolution issue
- has lots of data files and/or partitions
In this case, driver schema merging can be both slow and unnecessary. Would be good to have a configuration to let the use disable schema merging when converting such a metastore Parquet table.