Trapezoidal rule in log-log space

Consider the problem described in this StackOverFlow post. You have a function with certain smoothness properties that are apparent on a log-log plot. This is often accompanied by a large domain of integration. It seems worthwhile to "integrate in logspace", whatever that means. 

This Jupyter notebook probes this question and makes some recommendations.