This fix is to add one line to explain the current behavior of Spark SQL when writing Parquet files. All columns are forced to be nullable for compatibility reasons.

Author: gatorsmile <gatorsmile@gmail.com>

Closes #9314 from gatorsmile/lossNull.