[SPARK-30354][ML] GBT reuse DecisionTreeMetadata among iterations

### What changes were proposed in this pull request?
precompute the `DecisionTreeMetadata` and reuse it for all trees

### Why are the changes needed?
In existing impl, each `DecisionTreeRegressor` needs a pass on the whole dataset to calculate the same `DecisionTreeMetadata` repeatedly.
In this PR, with default depth=5, it is about 8% faster then existing impl

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

### How was this patch tested?
existing testsuites

Closes apache#27011 from zhengruifeng/gbt_reuse_instr_meta.

Authored-by: zhengruifeng <ruifengz@foxmail.com>
Signed-off-by: Sean Owen <srowen@gmail.com>

Author: zhengruifeng

mllib/src/main/scala/org/apache/spark/ml/tree/impl/GradientBoostedTrees.scala

@@ -20,7 +20,7 @@ package org.apache.spark.ml.tree.impl
 import org.apache.spark.internal.Logging
 import org.apache.spark.ml.feature.Instance
 import org.apache.spark.ml.linalg.Vector
-import org.apache.spark.ml.regression.{DecisionTreeRegressionModel, DecisionTreeRegressor}
+import org.apache.spark.ml.regression.DecisionTreeRegressionModel
 import org.apache.spark.mllib.tree.configuration.{Algo => OldAlgo}
 import org.apache.spark.mllib.tree.configuration.{BoostingStrategy => OldBoostingStrategy}
 import org.apache.spark.mllib.tree.impurity.{Variance => OldVariance}
@@ -306,8 +306,12 @@ private[spark] object GradientBoostedTrees extends Logging {
 
     // Initialize tree
     timer.start("building tree 0")
-    val firstTree = new DecisionTreeRegressor().setSeed(seed)
-    val firstTreeModel = firstTree.train(input, treeStrategy, featureSubsetStrategy)
+    val metadata = RandomForest.buildMetadata(input, treeStrategy,
+      numTrees = 1, featureSubsetStrategy)
+    val firstTreeModel = RandomForest.run(input, treeStrategy, numTrees = 1,
+      featureSubsetStrategy, seed = seed, instr = None,
+      parentUID = None, precomputedMetadata = Some(metadata))
+      .head.asInstanceOf[DecisionTreeRegressionModel]
     val firstTreeWeight = 1.0
     baseLearners(0) = firstTreeModel
     baseLearnerWeights(0) = firstTreeWeight
@@ -342,8 +346,10 @@ private[spark] object GradientBoostedTrees extends Logging {
       logDebug("Gradient boosting tree iteration " + m)
       logDebug("###################################################")
 
-      val dt = new DecisionTreeRegressor().setSeed(seed + m)
-      val model = dt.train(data, treeStrategy, featureSubsetStrategy)
+      val model = RandomForest.run(data, treeStrategy, numTrees = 1,
+        featureSubsetStrategy, seed = seed + m, instr = None,
+        parentUID = None, precomputedMetadata = Some(metadata))
+        .head.asInstanceOf[DecisionTreeRegressionModel]
       timer.stop(s"building tree $m")
       // Update partial model
       baseLearners(m) = model

mllib/src/main/scala/org/apache/spark/ml/tree/impl/RandomForest.scala

@@ -99,6 +99,15 @@ private[spark] object RandomForest extends Logging with Serializable {
     run(instances, strategy, numTrees, featureSubsetStrategy, seed, None)
   }
 
+  def buildMetadata(
+      input: RDD[Instance],
+      strategy: OldStrategy,
+      numTrees: Int,
+      featureSubsetStrategy: String): DecisionTreeMetadata = {
+    val retaggedInput = input.retag(classOf[Instance])
+    DecisionTreeMetadata.buildMetadata(retaggedInput, strategy, numTrees, featureSubsetStrategy)
+  }
+
   /**
    * Train a random forest.
    *
@@ -113,7 +122,8 @@ private[spark] object RandomForest extends Logging with Serializable {
       seed: Long,
       instr: Option[Instrumentation],
       prune: Boolean = true, // exposed for testing only, real trees are always pruned
-      parentUID: Option[String] = None): Array[DecisionTreeModel] = {
+      parentUID: Option[String] = None,
+      precomputedMetadata: Option[DecisionTreeMetadata] = None): Array[DecisionTreeModel] = {
 
     val timer = new TimeTracker()
 
@@ -122,8 +132,9 @@ private[spark] object RandomForest extends Logging with Serializable {
     timer.start("init")
 
     val retaggedInput = input.retag(classOf[Instance])
-    val metadata =
+    val metadata = precomputedMetadata.getOrElse {
       DecisionTreeMetadata.buildMetadata(retaggedInput, strategy, numTrees, featureSubsetStrategy)
+    }
 
     instr match {
       case Some(instrumentation) =>