Model prediction does not work in Python 3

[srikris]

With Python 3, the prediction code path using PyBind11 doesn't quite work. This is required to have full functionality with Python 3

Here is a simle repro step (this works in Python 2.7)
``
from sklearn import datasets
from sklearn.ensemble import RandomForestClassifier
import coremltools
import numpy

Train a scikit-learn model for Iris data

iris = datasets.load_iris()
clf = RandomForestClassifier()
clf.fit(iris.data, iris.target_names[iris.target])

Convert the model

model = coremltools.converters.sklearn.convert(clf)

Make predictions

model.predict({'input': np.zeros(4)})
``

[luke14free]

Not even conversion works. Some files make use of the xrange function

[TobyRoseman]

@luke14free - I currently have a pull request. Once that is merged then conversion should work.

[TobyRoseman]

@luke14free - the pull request has been merged. Please try conversion now in Python 3 and let us know if you run into any issues.