Zaporter/timeseries

[zaporter]

This adds support for time series into the optimizer.

Important things to know: *It currently just uses predicted_value to find utility. The error bounds on the predicted_value should be intelligently included into the calculations. *It has no unit tests. *It has mixed performance and has a hard time learning things like (ax^2)sin(bx) where x is the modulated variable. *It has a pretty clean interface and doesn't change the current API in any serious way. This might need to be altered a bit though *It only supports single objective optimization. We will require multi-objective optimization but this will require a much more complicated class that I will work on as soon as I am happy with the single-objective optimization