integrating robust covariances into the model fit method, see issue #1418
summary issue for robust covariances is #1158
first part: linear models - roughly finished
needed new df_resid_inference to separate df_resid used in scale from df_resid used in t, f tests and similar inference.
ENH: add cov_type to Regression model.fit(), first version
DOC: covtype in fit: update example with cluster
BUG: use use_t argument
REF: base Model, new optional attribute df_resid_inference
TST: robustcov add test for fit options and for scale, rsquared
Coverage increased (+0.03%) when pulling b20d35a on josef-pkt:REF_covtype_fit into bdb15bb on statsmodels:master.
I'm merging this.
Followup will be with next stage, robust covariance for discrete and general MLE.
GLM now also has score_obs, so I might be able to add it also.