TabularLime errors

I'm getting the following errors when using TabularLime:
"Output column features already exists."

Based off the [tests ](https://github.com/Azure/mmlspark/blob/99d267a407c07301dcada42f788ef7b92ef15aa5/src/image-featurizer/src/test/scala/LIMESuite.scala) for TabularLime, I should be calling a "fit" on TabularLime and then a "transform".

I'm on Spark 2.3.0 (Cloudera). The following shows my code to replicate this error:

```python
from pyspark.sql import SparkSession
from pyspark.ml.feature import VectorAssembler
from pyspark.ml import Pipeline
from pyspark.ml.classification import RandomForestClassifier

spark = SparkSession\
.builder\
.appName("pyspark")\
.getOrCreate()

from mmlspark import *
from mmlspark import FluentAPI

def generateDataFrame(spark):
data=[
{'x1':3, 'x2':4, 'y':1},
{'x1':2, 'x2':1, 'y':0},
{'x1':2, 'x2':2, 'y':0},
{'x1':3, 'x2':5, 'y':1}
]
return spark.createDataFrame(data)

vectorAssembler = VectorAssembler()\
.setInputCols(["x1", "x2"])\
.setOutputCol("features")

model=RandomForestClassifier(featuresCol="features", labelCol="y")

pipeline = Pipeline(stages=[
vectorAssembler, model
])

data=generateDataFrame(spark)

fitmodel=pipeline.fit(data)

lime = TabularLIME()\
.setModel(fitmodel)\
.setPredictionCol("prediction")\
.setOutputCol("weights")\
.setInputCol("features")

result = lime.fit(fitmodel.transform(data))

#This is where the error occurs
result.transform(fitmodel.transform(data))

spark.stop()
```

Start with the TabularLIME fit/transform flow in the linked LIMESuite.scala test and reproduce the reported error using the provided PySpark example on Spark 2.3.0. Check how the existing features and prediction columns are handled across fit and transform. Done means the example completes without the "Output column features already exists" error and the relevant test covers the behavior.

Written by the indexing model from the issue text.