Saves data directly from the database to a Parquet file using
DuckDB's native COPY command. This is much faster than reading
into R first and creates smaller files for easy sharing. Uses the data table name
stored in the connection attributes if available.