[SPARK-23604][SQL] Change Statistics.isEmpty to !Statistics.hasNonNul…

…lValue

## What changes were proposed in this pull request?

Parquet 1.9 will change the semantics of Statistics.isEmpty slightly
to reflect if the null value count has been set. That breaks a
timestamp interoperability test that cares only about whether there
are column values present in the statistics of a written file for an
INT96 column. Fix by using Statistics.hasNonNullValue instead.

## How was this patch tested?

Unit tests continue to pass against Parquet 1.8, and also pass against
a Parquet build including PARQUET-1217.

Author: Henry Robinson <henry@cloudera.com>

Closes #20740 from henryr/spark-23604.

sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/parquet/ParquetInteroperabilitySuite.scala

@@ -184,7 +184,7 @@ class ParquetInteroperabilitySuite extends ParquetCompatibilityTest with SharedS
                 // when the data is read back as mentioned above, b/c int96 is unsigned.  This
                 // assert makes sure this holds even if we change parquet versions (if eg. there
                 // were ever statistics even on unsigned columns).
-                assert(columnStats.isEmpty)
+                assert(!columnStats.hasNonNullValue)
               }
 
               // These queries should return the entire dataset with the conversion applied,