[SPARK-17182][SQL] Mark Collect as non-deterministic

## What changes were proposed in this pull request?

This PR marks the abstract class `Collect` as non-deterministic since the results of `CollectList` and `CollectSet` depend on the actual order of input rows.

## How was this patch tested?

Existing test cases should be enough.

Author: Cheng Lian <lian@databricks.com>

Closes #14749 from liancheng/spark-17182-non-deterministic-collect.

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/expressions/aggregate/collect.scala

@@ -54,6 +54,10 @@ abstract class Collect extends ImperativeAggregate {
 
   override def inputAggBufferAttributes: Seq[AttributeReference] = Nil
 
+  // Both `CollectList` and `CollectSet` are non-deterministic since their results depend on the
+  // actual order of input rows.
+  override def deterministic: Boolean = false
+
   protected[this] val buffer: Growable[Any] with Iterable[Any]
 
   override def initialize(b: MutableRow): Unit = {