ENH: design_info for endog

[josef-pkt]

patsy also provides a design_info for endog, that we currently don't attach and don't use.

This can be useful to look at possible endog transformation, or endog names in case of 2-D endog.

y, X = build_design_matrices([y.design_info. X.design_info], newdata)
https://groups.google.com/d/msg/pystatsmodels/sDED93s5Pk4/Ktf9QpPMAQAJ

related issues
#1449 endog transformation in predict
#2733 endog in Logit and GLM-Binomial treated as categorical by patsy, and a few more similar to this #878, #2181

[josef-pkt]

see also #7021 (comment)

I'd like patsy to leave endog alone. the model.__init__ should get the original endog except for dropping observations that have missing values in data for exog.