add discusison paragraph 2

[elray1]

content referring back to how our methods/results reinforce and are consistent with other findings, building an evidence base about success of ensembles, gradient boosting, transfer learning, careful feature engineering, data transformations.

[elray1]

I drafted some text along these lines but I felt that it actually made the paper seem weaker. It felt like we were saying, "we have no new results here; every modeling decision we made has been done before." I'm going to just close this issue and say the discussion is good enough.