Enhancement request: Add option to use robust regression to the linear fits.

[lukefullard]

Thank you for a super useful package. I have a suggestion for what I think is a simple enhancement.
I am plotting a periodogram, log(psd) vs log(frequency) and often there are spikes in the plot where there is a significant signal.
These "outliers" can cause the model fit to be biased and I would like to have the option to use robust regression to improve the fit.

Statsmodels has a pretty simple API for this and it is similar to the OLS class.
https://www.statsmodels.org/stable/rlm.html

I think it would be replacing
results = sm.OLS(endog=self.yy, exog=Z).fit()

with something similar to
results= sm.RLM(endog=self.yy, exog=Z, M=sm.robust.norms.HuberT()).fit()

You may also need to check that the output classes but from first glance these look compatible.

Thank you for your consideration.