Reduce duplication in the Cython/C++ for sandwich and dot products.

[ElizabethSantorellaQC]

Copied from https://github.com/Quantco/quantcore.glm/issues/221

There are several places where there's a lot of duplication: dense.pyx, dense_helpers-tmpl.cpp, categorical.pyx.

I'm not sure how feasible or reasonable it is to reduce this. It's going to be hard to do so without impacting performance.

[tbenthompson]

See #107 and #108

[jjerphan]

Hi, are you accepting external contributions?

[xhochy]

@jjerphan Yes, we are happy to receive external contributions.

[jjerphan]

Thanks for the information.

To me, it looks like the duplication in the files that are mentioned hardly can be removed without introducing extra complexity.

Yet, I think a few things might be simplified or improved in the code-base.

For now, I started with https://github.com/Quantco/tabmat/pull/245, a super minimalistic PR.