[SPARK-29118][ML] Avoid redundant computation in transform of GMM & GLR

### What changes were proposed in this pull request?
1,GMM: obtaining the prediction (double) from its probabilty prediction(vector)
2,GLR: obtaining the prediction (double) from its link prediction(double)

### Why are the changes needed?
it avoid predict twice

### Does this PR introduce any user-facing change?
no

### How was this patch tested?
existing tests

Closes #25815 from zhengruifeng/gmm_transform_opt.

Authored-by: zhengruifeng <ruifengz@foxmail.com>
Signed-off-by: Sean Owen <sean.owen@databricks.com>

Author: zhengruifeng

mllib/src/main/scala/org/apache/spark/ml/clustering/GaussianMixture.scala

@@ -33,8 +33,8 @@ import org.apache.spark.ml.util.Instrumentation.instrumented
 import org.apache.spark.mllib.linalg.{Matrices => OldMatrices, Matrix => OldMatrix,
   Vector => OldVector, Vectors => OldVectors}
 import org.apache.spark.rdd.RDD
-import org.apache.spark.sql.{Column, DataFrame, Dataset, Row, SparkSession}
-import org.apache.spark.sql.functions.udf
+import org.apache.spark.sql.{DataFrame, Dataset, Row, SparkSession}
+import org.apache.spark.sql.functions.{col, udf}
 import org.apache.spark.sql.types.{IntegerType, StructType}
 import org.apache.spark.storage.StorageLevel
 
@@ -111,28 +111,32 @@ class GaussianMixtureModel private[ml] (
   override def transform(dataset: Dataset[_]): DataFrame = {
     transformSchema(dataset.schema, logging = true)
 
-    var predictionColNames = Seq.empty[String]
-    var predictionColumns = Seq.empty[Column]
-
-    if ($(predictionCol).nonEmpty) {
-      val predUDF = udf((vector: Vector) => predict(vector))
-      predictionColNames :+= $(predictionCol)
-      predictionColumns :+= predUDF(DatasetUtils.columnToVector(dataset, getFeaturesCol))
-    }
+    val vectorCol = DatasetUtils.columnToVector(dataset, $(featuresCol))
+    var outputData = dataset
+    var numColsOutput = 0
 
     if ($(probabilityCol).nonEmpty) {
       val probUDF = udf((vector: Vector) => predictProbability(vector))
-      predictionColNames :+= $(probabilityCol)
-      predictionColumns :+= probUDF(DatasetUtils.columnToVector(dataset, getFeaturesCol))
+      outputData = outputData.withColumn($(probabilityCol), probUDF(vectorCol))
+      numColsOutput += 1
+    }
+
+    if ($(predictionCol).nonEmpty) {
+      if ($(probabilityCol).nonEmpty) {
+        val predUDF = udf((vector: Vector) => vector.argmax)
+        outputData = outputData.withColumn($(predictionCol), predUDF(col($(probabilityCol))))
+      } else {
+        val predUDF = udf((vector: Vector) => predict(vector))
+        outputData = outputData.withColumn($(predictionCol), predUDF(vectorCol))
+      }
+      numColsOutput += 1
     }
 
-    if (predictionColNames.nonEmpty) {
-      dataset.withColumns(predictionColNames, predictionColumns)
-    } else {
+    if (numColsOutput == 0) {
       this.logWarning(s"$uid: GaussianMixtureModel.transform() does nothing" +
         " because no output columns were set.")
-      dataset.toDF()
     }
+    outputData.toDF
   }
 
   @Since("2.0.0")

mllib/src/main/scala/org/apache/spark/ml/regression/GeneralizedLinearRegression.scala

@@ -1036,31 +1036,33 @@ class GeneralizedLinearRegressionModel private[ml] (
   }
 
   override protected def transformImpl(dataset: Dataset[_]): DataFrame = {
-    var predictionColNames = Seq.empty[String]
-    var predictionColumns = Seq.empty[Column]
-
     val offset = if (!hasOffsetCol) lit(0.0) else col($(offsetCol)).cast(DoubleType)
+    var outputData = dataset
+    var numColsOutput = 0
 
-    if ($(predictionCol).nonEmpty) {
-      val predictUDF = udf { (features: Vector, offset: Double) => predict(features, offset) }
-      predictionColNames :+= $(predictionCol)
-      predictionColumns :+= predictUDF(col($(featuresCol)), offset)
+    if (hasLinkPredictionCol) {
+      val predLinkUDF = udf((features: Vector, offset: Double) => predictLink(features, offset))
+      outputData = outputData
+        .withColumn($(linkPredictionCol), predLinkUDF(col($(featuresCol)), offset))
+      numColsOutput += 1
     }
 
-    if (hasLinkPredictionCol) {
-      val predictLinkUDF =
-        udf { (features: Vector, offset: Double) => predictLink(features, offset) }
-      predictionColNames :+= $(linkPredictionCol)
-      predictionColumns :+= predictLinkUDF(col($(featuresCol)), offset)
+    if ($(predictionCol).nonEmpty) {
+      if (hasLinkPredictionCol) {
+        val predUDF = udf((eta: Double) => familyAndLink.fitted(eta))
+        outputData = outputData.withColumn($(predictionCol), predUDF(col($(linkPredictionCol))))
+      } else {
+        val predUDF = udf((features: Vector, offset: Double) => predict(features, offset))
+        outputData = outputData.withColumn($(predictionCol), predUDF(col($(featuresCol)), offset))
+      }
+      numColsOutput += 1
     }
 
-    if (predictionColNames.nonEmpty) {
-      dataset.withColumns(predictionColNames, predictionColumns)
-    } else {
+    if (numColsOutput == 0) {
       this.logWarning(s"$uid: GeneralizedLinearRegressionModel.transform() does nothing" +
         " because no output columns were set.")
-      dataset.toDF()
     }
+    outputData.toDF
   }
 
   /**