Shoulders of Giants: A Look at the Degree and Utility of Openness in NLP Research