java implementation of LightGBM predicting part

This project is a java translation of LightGBM predicting part at version 57d552726f19ba9b29843cbaa50f23a3ba4e79e6 in master.

LightGBM is an excellent tool for training model in offline way. However, when a model is saved by LightGBM, it is difficult to be used efficiently for online predicting in production environment.

I translate the predicting part of LightGBM into java so that the model can be used for online predicting in java applications. After training a model with LightGBM, you can use the model to do predicting job for every realtime query data, just like the unit test code shows. Note that org.lightgbm.predict4j.v2.UseageTest is for LightGBM v2 while org.lightgbm.predict4j.UseageTest is for LightGBM v1.

2017-07-04: Add supports for LightGBM v2 (version d0858b36765191899d0a47ba1a52bd70e78dad22)