ENH: relationship eta-squared effect size and semi-partial correlation

[josef-pkt]

https://github.com/friendly/heplots/issues/1

"
I suspect DescTools is reporting the correct results since $\eta^2$ is really just the semi-partial correlation squared, which can be obtained via R^2 differences after removing the IV of interest.
"

We get semi-partial correlation in PR #2380 (at least I wrote the functions in that context)
and get etasq effect size in #6526 for anova case, and open issue #5896

matrix version of semi-partial correlation only works for single columns and not for multi-column categorical variables as in anova.

[josef-pkt]

related: confidence intervals

I'm adding confidence intervals for effect size estimates. There is also a literature for confidence intervals and inference for correlation and partial correlation.

I have no overview yet what is the same and can be combined.