[SPARK-48544][SQL] Reduce memory pressure of empty TreeNode BitSets

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

- Changed the `ineffectiveRules` variable of the `TreeNode` class to initialize lazily. This will reduce unnecessary driver memory pressure.

### Why are the changes needed?

- Plans with large expression or operator trees are known to cause driver memory pressure; this is one step in alleviating that issue.

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

No.

### How was this patch tested?

Existing UT covers behavior. Outwards facing behavior does not change.

### Was this patch authored or co-authored using generative AI tooling?

No

Closes #46919 from n-young-db/ineffective-rules-lazy.

Authored-by: Nick Young <nick.young@databricks.com>
Signed-off-by: Josh Rosen <joshrosen@databricks.com>

Author: n-young-db

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/trees/TreeNode.scala

@@ -120,7 +120,14 @@ abstract class TreeNode[BaseType <: TreeNode[BaseType]]
    * ineffective for subsequent apply calls on this tree because query plan structures are
    * immutable.
    */
-  private val ineffectiveRules: BitSet = new BitSet(RuleIdCollection.NumRules)
+  private[this] var _ineffectiveRules: BitSet = null
+  private def ineffectiveRules: BitSet = {
+    if (_ineffectiveRules eq null) {
+      _ineffectiveRules = new BitSet(RuleIdCollection.NumRules)
+    }
+    _ineffectiveRules
+  }
+  private def isIneffectiveRulesEmpty = _ineffectiveRules eq null
 
   /**
    * @return a sequence of tree pattern enums in a TreeNode T. It does not include propagated
@@ -149,7 +156,7 @@ abstract class TreeNode[BaseType <: TreeNode[BaseType]]
    *         UnknownId, it returns false.
    */
   protected def isRuleIneffective(ruleId : RuleId): Boolean = {
-    if (ruleId eq UnknownRuleId) {
+    if (isIneffectiveRulesEmpty || (ruleId eq UnknownRuleId)) {
       return false
     }
     ineffectiveRules.get(ruleId.id)