Adding the Keith et al. 2013 DM Modeling to PINT

[gooddc]

Implement the DM modeling strategy outlined in Keith et al. 2013 to PINT.

The motivation here is to provide continuity with other PTA and IPTA analysis methods. This doesn't add much new vs. DMX for the NANOGrav side, but provides a stopgap for IPTA DR3 as we work towards advanced noise modeling/Gaussian Process modeling and increases the amount of non-NANOGrav PTA techniques we can replicate.

IPTA DR2 Version A used this DM modeling. That means that the release includes tempo2 based par/tim files using the Keith et al. 2013 DM Modeling, and it's accessible here. These can be used as a double check, by running through tempo2 then comparing to PINT output.

[aarchiba]

If I've understood the paper correctly on a quick skim, the only low-level machinery we'd need would be to allow DMX to do linear interpolation rather than piecewise constant. At a higher level we'd have a loop that redesigned the DMX parameters based on fit results. I think the former could be controlled by a parfile switch (say DMX_LINEAR Y), and the latter would just be a script that calls PINT stuff. Or one could make a Fitter, I suppose.