[pystatsmodels] Design: more tool boxes or making helper functions reusable

[j***@gmail.com]

Whenever I work on a new model myself, then I wish we had more tool
functions that are discoverable and reusable. We have many helper
functions in various models that were written for the specific
usecase. However, those are difficult to find and might not be general
enough and might not have a docstring that is informative enough for
reuse.

I also ended up writing similar function several times which can take
half a minute or a few hours.

When we implement a new model it can be distracting to think about and
work on reusability of helper functions. However, every once in a
while we should spend some effort making those helper functions
reusable because it will save time (and have already unit tests) the
next time we need them for a different model.

linalg tools
https://github.com/statsmodels/statsmodels/issues/4992

binning matching tools
https://github.com/statsmodels/statsmodels/issues/5388

matrix (cov) tools
https://github.com/statsmodels/statsmodels/pull/4143
https://github.com/statsmodels/statsmodels/issues/4140

(There are also pca and cca functions for when we don't need the full
model classes.)

We already have quite a lot of tool functions, but we need more.
We also have many methods in the model and results classes that could
be reusable functions for cases that inheritance does not cover.

Josef