Order testing

[ChrisRackauckas]

We need a way to test that the order of accuracy is correct.

[ChrisRackauckas]

Here's how this should be done:

• Pick a function f(t) and a starting set of time points t1 which is unevenly spaced.
• At 100 evenly spaced points (make sure they aren't on the grid of known values), calculate the mean squared error between the interpolant and f(t).
• Refine to t2 which is t1 and each half point.
• Calculate the error at the same 100 evenly spaced points
• Repeat N times (user chosen)
• Use the log approximation to determine the order of convergence. Return a ConvergenceSimulation from DiffEqDevTools.

This should probably just be ConvergenceSimulation so that all of the plots and such are done. This error estimator would be an L2 error approximation.