[SPARK-43885][SQL][FOLLOWUP] Instruction#dataType should not fail

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

This is a followup of apache#41448 . As an optimizer rule, the produced plan should be resolved and resolved expressions should be able to report data type. The `Instruction` expression fails to report data type and may break external optimizer rules. This PR makes it to return dummy NullType.

### Why are the changes needed?

to not break external optimizer rules.

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

no

### How was this patch tested?

existing tests

Closes apache#42482 from cloud-fan/merge.

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/plans/logical/MergeRows.scala

@@ -21,7 +21,7 @@ import org.apache.spark.sql.catalyst.expressions.{Attribute, AttributeSet, Expre
 import org.apache.spark.sql.catalyst.plans.logical.MergeRows.{Instruction, ROW_ID}
 import org.apache.spark.sql.catalyst.trees.UnaryLike
 import org.apache.spark.sql.catalyst.util.truncatedString
-import org.apache.spark.sql.types.DataType
+import org.apache.spark.sql.types.{DataType, NullType}
 
 case class MergeRows(
     isSourceRowPresent: Expression,
@@ -74,7 +74,11 @@ object MergeRows {
     def condition: Expression
     def outputs: Seq[Seq[Expression]]
     override def nullable: Boolean = false
-    override def dataType: DataType = throw new UnsupportedOperationException("dataType")
+    // We return NullType here as only the `MergeRows` operator can contain `Instruction`
+    // expressions and it doesn't care about the data type. Some external optimizer rules may
+    // assume optimized plan is always resolved and Expression#dataType is always available, so
+    // we can't just fail here.
+    override def dataType: DataType = NullType
   }
 
   case class Keep(condition: Expression, output: Seq[Expression]) extends Instruction {