Xgboost models support

[dmitryikh]

for #2 add gbtree support:

• load models from binary format
• perform predictions on:
  • single feature vector
  • dense matrix of feature vectors
  • sparse (CSR) matrix of feature vectors
• do not support trasformation function (predictions are raw scores)
• add test on agaricus data

[coveralls]

Pull Request Test Coverage Report for Build 22

• 211 of 332 (63.55%) changed or added relevant lines in 4 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+4.9%) to 64.773%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
xgensemble.go	31	42	73.81%
internal/xgbin/xgbin_io.go	64	119	53.78%
xgensemble_io.go	109	164	66.46%
<!--	Total:	211	332	63.55%	-->

Totals
Change from base Build 16:	4.9%
Covered Lines:	627
Relevant Lines:	968

---

💛 - Coveralls