[SPARK-48155][FOLLOWUP][SQL] AQEPropagateEmptyRelation for left anti …

…join should check if remain child is just BroadcastQueryStageExec

### What changes were proposed in this pull request?
As title.

### Why are the changes needed?

We encountered BroadcastNestedLoopJoin LeftAnti BuildLeft, and it's right is empty. It is left child of left outer BroadcastHashJoin. The case is more complicated, part of the Initial Plan is as follows
```
               :- Project (214)
               :  +- BroadcastHashJoin LeftOuter BuildRight (213)
               :     :- BroadcastNestedLoopJoin LeftAnti BuildLeft (211)
               :     :  :- BroadcastExchange (187)
               :     :  :  +- Project (186)
               :     :  :     +- Filter (185)
               :     :  :        +- Scan parquet  (31)
               :     :  +- LocalLimit (210)
               :     :     +- Project (209)
               :     :        +- BroadcastHashJoin Inner BuildLeft (208)
               :     :           :- BroadcastExchange (194)
               :     :           :  +- Project (193)
               :     :           :     +- BroadcastHashJoin LeftOuter BuildRight (192)
               :     :           :        :- Project (189)
               :     :           :        :  +- Filter (188)
               :     :           :        :     +- Scan parquet  (37)
               :     :           :        +- BroadcastExchange (191)
```
After AQEPropagateEmptyRelation, report an error "HashJoin should not take LeftOuter as the JoinType with building left side"

### Does this PR introduce _any_ user-facing change?
No.

### How was this patch tested?

### Was this patch authored or co-authored using generative AI tooling?
No.

Closes #48300 from zml1206/SPARK-48155-followup.

Authored-by: zml1206 <zhuml1206@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>

sql/catalyst/src/main/scala/org/apache/spark/sql/catalyst/optimizer/PropagateEmptyRelation.scala

@@ -111,7 +111,8 @@ abstract class PropagateEmptyRelationBase extends Rule[LogicalPlan] with CastSup
           // Except is handled as LeftAnti by `ReplaceExceptWithAntiJoin` rule.
           case LeftOuter | LeftSemi | LeftAnti if isLeftEmpty => empty(p)
           case LeftSemi if isRightEmpty | isFalseCondition => empty(p)
-          case LeftAnti if isRightEmpty | isFalseCondition => p.left
+          case LeftAnti if (isRightEmpty | isFalseCondition) && canExecuteWithoutJoin(p.left) =>
+            p.left
           case FullOuter if isLeftEmpty && isRightEmpty => empty(p)
           case LeftOuter | FullOuter if isRightEmpty && canExecuteWithoutJoin(p.left) =>
             Project(p.left.output ++ nullValueProjectList(p.right), p.left)

sql/core/src/test/scala/org/apache/spark/sql/execution/adaptive/AdaptiveQueryExecSuite.scala

@@ -2829,6 +2829,38 @@ class AdaptiveQueryExecSuite
       assert(findTopLevelBroadcastNestedLoopJoin(adaptivePlan).size == 1)
       assert(findTopLevelUnion(adaptivePlan).size == 0)
     }
+
+    withSQLConf(
+      SQLConf.AUTO_BROADCASTJOIN_THRESHOLD.key -> "100") {
+      withTempView("t1", "t2", "t3", "t4") {
+        Seq(1).toDF().createOrReplaceTempView("t1")
+        spark.range(100).createOrReplaceTempView("t2")
+        spark.range(2).createOrReplaceTempView("t3")
+        spark.range(2).createOrReplaceTempView("t4")
+        val (_, adaptivePlan) = runAdaptiveAndVerifyResult(
+          """
+            |SELECT tt2.value
+            |FROM (
+            |  SELECT value
+            |  FROM t1
+            |  WHERE NOT EXISTS (
+            |      SELECT 1
+            |      FROM (
+            |        SELECT t2.id
+            |        FROM t2
+            |          JOIN t3 ON t2.id = t3.id
+            |        AND t2.id > 100
+            |      ) tt
+            |      WHERE t1.value = tt.id
+            |    )
+            |    AND t1.value = 1
+            |) tt2
+            |  LEFT JOIN t4 ON tt2.value = t4.id
+            |""".stripMargin
+        )
+        assert(findTopLevelBroadcastNestedLoopJoin(adaptivePlan).size == 1)
+      }
+    }
   }
 
   test("SPARK-39915: Dataset.repartition(N) may not create N partitions") {