Add partial dependency subclass for text

[aikramer2]

[aikramer2]

we should enable this pattern:

docs = ['this is a sentence','so is this']
processor = FunctionTransformer(func = func)
pipeline = make_pipeline(processor, model)
i = TextInterpretation()
i.load_data(docs, type = 'text', bag_of_words = True, tokenizer = None)
[generate count vectorizer]
i.partial_dependence(word, pipeline.predict_proba)