[SPARK-2478] [mllib] DecisionTree Python API

[jkbradley]

Added experimental Python API for Decision Trees.

API:

• class DecisionTreeModel
  ** predict() for single examples and RDDs, taking both feature vectors and LabeledPoints
  ** numNodes()
  ** depth()
  ** str()
• class DecisionTree
  ** trainClassifier()
  ** trainRegressor()
  ** train()

Examples and testing:

• Added example testing classification and regression with batch prediction: examples/src/main/python/mllib/tree.py
• Have also tested example usage in doc of python/pyspark/mllib/tree.py which tests single-example prediction with dense and sparse vectors

Also: Small bug fix in python/pyspark/mllib/_common.py: In _linear_predictor_typecheck, changed check for RDD to use isinstance() instead of type() in order to catch RDD subclasses.

CC @mengxr @manishamde

[jkbradley]

@mengxr Hopefully good to go if Jenkins agrees.

[SparkQA]

QA tests have started for PR 1727. This patch merges cleanly.
View progress: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/17758/consoleFull

[SparkQA]

QA results for PR 1727:
- This patch PASSES unit tests.
- This patch merges cleanly
- This patch adds the following public classes (experimental):
Re-index class labels in a dataset to the range {0,...,numClasses-1}.
class DecisionTreeModel(object):
class DecisionTree(object):

For more information see test ouptut:
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/17758/consoleFull

[manishamde]

👍 Not very familiar with the python API but the DT changes look good to me. I will be happy to test this for QA. :-)

[mengxr]

minor: Do we want to check data.count() == 0?

[SparkQA]

QA tests have started for PR 1727. This patch merges cleanly.
View progress: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/17773/consoleFull

[SparkQA]

QA results for PR 1727:
- This patch PASSES unit tests.
- This patch merges cleanly
- This patch adds the following public classes (experimental):
Re-index class labels in a dataset to the range {0,...,numClasses-1}.
class DecisionTreeModel(object):
class DecisionTree(object):

For more information see test ouptut:
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/17773/consoleFull

[mengxr]

LGTM. Merged into both master and branch-1.1. Thanks!!