fix: ignore reader-side parquet.hadoop.vectored.io.enabled in Iceberg native-write detection

spark/src/main/scala/org/apache/comet/serde/operator/CometIcebergNativeWrite.scala

@@ -89,6 +89,16 @@ object CometIcebergNativeWrite extends CometOperatorSerde[IcebergWriteExec] with
   private val ParquetWritePropertyPrefix = "write.parquet."
   private val ParquetMrPropertyPrefix = "parquet."
 
+  // Hadoop-side `parquet.*` keys that iceberg-java's writer never consumes, so seeing them
+  // in the session Hadoop configuration does not indicate the native writer would diverge.
+  // `parquet.hadoop.vectored.io.enabled` is a reader-side vectored-IO knob declared by
+  // parquet-hadoop as `ParquetInputFormat.HADOOP_VECTORED_IO_ENABLED` (default `true` in
+  // parquet-hadoop 1.16+) and only consulted by parquet-mr's Hadoop reader path. Keep it
+  // out of the writer-compatibility gate so that environments which seed it into the
+  // session Hadoop configuration do not silently disable native Iceberg writes.
+  private val IgnoredHadoopParquetConfKeys: Set[String] = Set(
+    "parquet.hadoop.vectored.io.enabled")
+
   private lazy val vettedParquetWriteKeys: Set[String] = Set(
     PropertyKeys.ParquetCompressionCodec,
     PropertyKeys.ParquetCompressionLevel,
@@ -271,7 +281,8 @@ object CometIcebergNativeWrite extends CometOperatorSerde[IcebergWriteExec] with
   private val requireNoParquetHadoopConfOverrides: TriggerRule = ctx =>
     ctx.hadoopConf.asScala
       .map(_.getKey)
-      .find(_.startsWith(ParquetMrPropertyPrefix))
+      .filter(_.startsWith(ParquetMrPropertyPrefix))
+      .find(k => !IgnoredHadoopParquetConfKeys.contains(k))
       .map(k => s"Hadoop configuration sets $k (reaches iceberg-java's writer but not native)")
 
   private val requireSupportedStorageScheme: TriggerRule = ctx =>

spark/src/test/scala/org/apache/comet/CometIcebergWriteDetectionSuite.scala

@@ -357,6 +357,20 @@ class CometIcebergWriteDetectionSuite extends CometTestBase with CometIcebergTes
     }
   }
 
+  // parquet.hadoop.vectored.io.enabled is a reader-side vectored-IO knob declared by
+  // parquet-hadoop (ParquetInputFormat.HADOOP_VECTORED_IO_ENABLED, default true in
+  // parquet-hadoop 1.16+). iceberg-java's writer never consumes it, so it must not
+  // disable native Iceberg writes when it happens to be present in the session
+  // Hadoop configuration.
+  test("Compatible when only parquet.hadoop.vectored.io.enabled is set in Hadoop configuration") {
+    withDetectionCatalog { dir =>
+      createTable(dir, "vectored_io_only", partitionSpec = "")
+      withSQLConf("parquet.hadoop.vectored.io.enabled" -> "true") {
+        assertSupportLevelIs[Compatible]("vectored_io_only")
+      }
+    }
+  }
+
   test("fall-back: io-impl set") {
     withDetectionCatalog { dir =>
       createTable(