GPU binning and compression. (#3319)

* GPU binning and compression.

- binning and index compression are done inside the DeviceShard constructor
- in case of a DMatrix with multiple row batches, it is first converted into a single row batch

tests/cpp/helpers.h

@@ -23,6 +23,8 @@ long GetFileSize(const std::string filename);
 
 std::string CreateSimpleTestData();
 
+std::string CreateBigTestData(size_t n_entries);
+
 void CheckObjFunction(xgboost::ObjFunction * obj,
                       std::vector<xgboost::bst_float> preds,
                       std::vector<xgboost::bst_float> labels,