[SQL] Don't shuffle code generated rows

When sort based shuffle and code gen are on we were trying to ship the code generated rows during a shuffle.  This doesn't work because the classes don't exist on the other side.  Instead we now copy into a generic row before shipping.

Author: Michael Armbrust <michael@databricks.com>

Closes #3263 from marmbrus/aggCodeGen and squashes the following commits:

f6ba8cf [Michael Armbrust] fix and test

sql/core/src/main/scala/org/apache/spark/sql/execution/Exchange.scala

@@ -47,8 +47,8 @@ case class Exchange(newPartitioning: Partitioning, child: SparkPlan) extends Una
         // TODO: Eliminate redundant expressions in grouping key and value.
         val rdd = if (sortBasedShuffleOn) {
           child.execute().mapPartitions { iter =>
-            val hashExpressions = newProjection(expressions, child.output)
-            iter.map(r => (hashExpressions(r), r.copy()))
+            val hashExpressions = newMutableProjection(expressions, child.output)()
+            iter.map(r => (hashExpressions(r).copy(), r.copy()))
           }
         } else {
           child.execute().mapPartitions { iter =>

sql/core/src/test/scala/org/apache/spark/sql/SQLQuerySuite.scala

@@ -72,6 +72,13 @@ class SQLQuerySuite extends QueryTest with BeforeAndAfterAll {
       2.5)
   }
 
+  test("aggregation with codegen") {
+    val originalValue = codegenEnabled
+    setConf(SQLConf.CODEGEN_ENABLED, "true")
+    sql("SELECT key FROM testData GROUP BY key").collect()
+    setConf(SQLConf.CODEGEN_ENABLED, originalValue.toString)
+  }
+
   test("SPARK-3176 Added Parser of SQL LAST()") {
     checkAnswer(
       sql("SELECT LAST(n) FROM lowerCaseData"),