Revert "[spark] Move row-level write rules into Resolution batch for Spark 4.1"

This reverts commit cc1d93c.

Author: kerwin-zk

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

@@ -47,18 +47,9 @@ class PaimonSparkSessionExtensions extends (SparkSessionExtensions => Unit) {
     extensions.injectPostHocResolutionRule(spark => ReplacePaimonFunctions(spark))
     extensions.injectPostHocResolutionRule(spark => PaimonPostHocResolutionRules(spark))
 
-    // Row-level write rewrites must fire during the Resolution batch, not as postHoc rules. Spark
-    // 4.1 moved `RewriteUpdateTable` / `RewriteDeleteFromTable` / `RewriteMergeIntoTable` into the
-    // Resolution batch and marks the plan as `analyzed=true` by the time postHoc rules run;
-    // `AnalysisHelper.resolveOperators*` then short-circuits and Spark 4.1 treats the postHoc
-    // rule's return value as a no-op, leaving the original UpdateTable node for the physical
-    // planner to reject with `UNSUPPORTED_FEATURE.TABLE_OPERATION`. Registering as resolution
-    // rules lets them transform before the analyzed flag is set. See
-    // AnalysisHelper.resolveOperatorsUpWithPruning: it returns `self` unchanged when
-    // `analyzed=true`, which is the root cause of the Spark 4.1 append-only UPDATE/DELETE failures.
-    extensions.injectResolutionRule(_ => PaimonUpdateTable)
-    extensions.injectResolutionRule(_ => PaimonDeleteTable)
-    extensions.injectResolutionRule(spark => PaimonMergeInto(spark))
+    extensions.injectPostHocResolutionRule(_ => PaimonUpdateTable)
+    extensions.injectPostHocResolutionRule(_ => PaimonDeleteTable)
+    extensions.injectPostHocResolutionRule(spark => PaimonMergeInto(spark))
 
     // table function extensions
     PaimonTableValuedFunctions.supportedFnNames.foreach {