[SPARK-30620][SQL] avoid unnecessary serialization in AggregateExpres…

…sion

### What changes were proposed in this pull request?

Expressions are very likely to be serialized and sent to executors, we should avoid unnecessary serialization overhead as much as we can.

This PR fixes `AggregateExpression`.

### Why are the changes needed?

small improvement

### Does this PR introduce any user-facing change?

no

### How was this patch tested?

existing tests

Closes #27342 from cloud-fan/fix.

Authored-by: Wenchen Fan <wenchen@databricks.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>

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

@@ -96,6 +96,7 @@ case class AggregateExpression(
   extends Expression
   with Unevaluable {
 
+  @transient
   lazy val resultAttribute: Attribute = if (aggregateFunction.resolved) {
     AttributeReference(
       aggregateFunction.toString,
@@ -108,7 +109,7 @@ case class AggregateExpression(
     UnresolvedAttribute(aggregateFunction.toString)
   }
 
-  lazy val filterAttributes: AttributeSet = filter.map(_.references).getOrElse(AttributeSet.empty)
+  def filterAttributes: AttributeSet = filter.map(_.references).getOrElse(AttributeSet.empty)
 
   // We compute the same thing regardless of our final result.
   override lazy val canonicalized: Expression = {