Update timestep dependence plot to extrapolate to zero timestep

[jchodera]

I think we should update Fig 4 (timestep dependence plot) with the following:

• Caption should begin with boldfaced sentence fragment summarizing plot main message
• x-axis should span [0,2.5] fs
• Panels (a) and (b) should be labeled (a) and (b). Alternatively, we can refer to them as "top" and "bottom".
• The top plot should have a linear interpolation to zero timestep; that should be our "reference value" for computing the relative error. Same with bottom plot.
• We should add 1.5 fs data point.

[kyleabeauchamp]

Fixed by #28