[spark] Revert MergeInto changes from Round 8; keep Update/Delete Resolution rule fix

Round 8 (72ef84b) fixed the 5 Spark 4.1 append-only UPDATE/DELETE failures
by moving PaimonUpdateTable / PaimonDeleteTable / PaimonMergeInto into the
Resolution batch. However, it regressed 3 Blob merge-into tests that were
previously passing via the postHoc MergeInto path.

The append-only UPDATE/DELETE root cause is specific to UpdateTable /
DeleteFromTable plan types on Spark 4.1 — Spark 4.1's analyzed-plan handling
silently drops postHoc transformed output for these specific node types.
MergeIntoTable does not have the same issue: the postHoc PaimonMergeInto rule
was correctly transforming MergeIntoTable -> MergeIntoPaimonDataEvolutionTable
/ MergeIntoPaimonTable for Blob tests before Round 8.

Keep PaimonUpdateTable / PaimonDeleteTable registered additionally as
Resolution rules on Spark 4.1+ (with the guard to skip SparkTableWithRowLevelOps
so Spark's V2 rewrite handles useV2Write=true). Restore PaimonMergeInto to the
postHoc-only registration (and drop the guard from PaimonMergeIntoBase) — this
matches the original Round 3 behavior that Blob tests depend on.

Author: kerwin-zk

paimon-spark/paimon-spark-common/src/main/scala/org/apache/paimon/spark/catalyst/analysis/PaimonMergeIntoBase.scala

@@ -18,7 +18,7 @@
 
 package org.apache.paimon.spark.catalyst.analysis
 
-import org.apache.paimon.spark.{SparkTable, SparkTableWithRowLevelOps}
+import org.apache.paimon.spark.SparkTable
 import org.apache.paimon.spark.catalyst.analysis.expressions.ExpressionHelper
 import org.apache.paimon.spark.commands.{MergeIntoPaimonDataEvolutionTable, MergeIntoPaimonTable}
 
@@ -47,16 +47,10 @@ trait PaimonMergeIntoBase
     // `transformDown` (which unconditionally visits every node) guarded by
     // `AnalysisHelper.allowInvokingTransformsInAnalyzer` so the in-analyzer assertion does not
     // trip.
-    val skipV2RowLevelTables = org.apache.spark.SPARK_VERSION >= "4.1"
     AnalysisHelper.allowInvokingTransformsInAnalyzer {
       plan.transformDown {
         case merge: MergeIntoTable
-            if merge.resolved && PaimonRelation.isPaimonTable(merge.targetTable) &&
-              (!skipV2RowLevelTables ||
-                !PaimonRelation
-                  .getPaimonRelation(merge.targetTable)
-                  .table
-                  .isInstanceOf[SparkTableWithRowLevelOps]) =>
+            if merge.resolved && PaimonRelation.isPaimonTable(merge.targetTable) =>
           val relation = PaimonRelation.getPaimonRelation(merge.targetTable)
           val v2Table = relation.table.asInstanceOf[SparkTable]
           val dataEvolutionEnabled = v2Table.coreOptions.dataEvolutionEnabled()

paimon-spark/paimon-spark-common/src/main/scala/org/apache/paimon/spark/extensions/PaimonSparkSessionExtensions.scala

@@ -51,19 +51,20 @@ class PaimonSparkSessionExtensions extends (SparkSessionExtensions => Unit) {
     extensions.injectPostHocResolutionRule(_ => PaimonDeleteTable)
     extensions.injectPostHocResolutionRule(spark => PaimonMergeInto(spark))
 
-    // Spark 4.1 additionally registers the same rules in the Resolution batch. Spark 4.1 moved its
-    // own RewriteUpdateTable / RewriteDeleteFromTable / RewriteMergeIntoTable into the Resolution
-    // batch and marks the plan as analyzed=true before postHoc rules run. The postHoc rule's
-    // transformed output is then silently dropped, leaving the original UpdateTable / DeleteFromTable
-    // / MergeIntoTable in the plan for the physical planner to reject with
-    // UNSUPPORTED_FEATURE.TABLE_OPERATION. Running them in the Resolution batch lets Paimon's
-    // rewrite actually stick. The rules themselves skip tables advertising SupportsRowLevelOperations
-    // (gated on the same Spark 4.1 version check), so Spark's built-in rewrite still owns the V2
-    // row-level path for useV2Write=true tables.
+    // Spark 4.1 additionally registers Update/Delete rules in the Resolution batch. Spark 4.1 moved
+    // its own RewriteUpdateTable / RewriteDeleteFromTable into the Resolution batch and marks the
+    // plan as analyzed=true before postHoc rules run. The postHoc rule's transformed output is then
+    // silently dropped for UpdateTable / DeleteFromTable nodes, leaving them in the plan for the
+    // physical planner to reject with UNSUPPORTED_FEATURE.TABLE_OPERATION. Running these rules in
+    // the Resolution batch lets Paimon's rewrite actually stick. Each rule skips tables advertising
+    // SupportsRowLevelOperations (gated on the same Spark 4.1 version check) so Spark's built-in
+    // rewrite owns the V2 row-level path for useV2Write=true tables. `PaimonMergeInto` is
+    // intentionally left out here — the postHoc path already works for merge plans (verified by
+    // Blob merge tests passing pre-Round-8 on PR 7648) and re-registering it in the Resolution
+    // batch introduced regressions on Blob tests.
     if (org.apache.spark.SPARK_VERSION >= "4.1") {
       extensions.injectResolutionRule(_ => PaimonUpdateTable)
       extensions.injectResolutionRule(_ => PaimonDeleteTable)
-      extensions.injectResolutionRule(spark => PaimonMergeInto(spark))
     }
 
     // table function extensions