PINT does not easily provide whitened residuals

[aarchiba]

If one has ECORR or red noise, there are two notions of residual - one that is the plain residuals, after subtraction of the timing model, and one that has been "whitened" by subtraction of the basis functions corresponding to those noise components. One could quite reasonably want to plot either, but PINT provides only the former. GLSFitter and WidebandTOAFitter do add a .noise_resids dictionary to their resids object, but it is not trivial to get whitened residuals from this (and it's not at all clear how this behaves for WidebandTOAFitter). Perhaps a .whitened attribute or property on appropriate Residuals objects? Or Fitters could offer, if relevant, a .whitened_resids attribute alongside their .resids?

Convenient access to whitened residuals would be valuable for, e.g., identifying outliers for manual TOA excision, or for visually inspecting fits for plausibility.

[aarchiba]

https://gitlab.nanograv.org/nano-time/timing_analysis/blob/15yr/src/timing_analysis/utils.py#L20 is a demonstration (though wrong) of how to do it