Support text format in classification model and clustering model

Currently, these models support only non-text format.
This bp aims to support text format using tf–idf. [1]

A tf–idf is a text-mining technology which parse documents to index by numerical statistic.

This feature allows user to create following prediction models.

- model detects whether it is a spam mail or not
- model predicts whether it is a review of goodwill or not
- model detects what language a document is written in

[1] https://spark.apache.org/docs/1.6.3/mllib-feature-extraction.html#tf-idf