Proposal: add vcov `attr`s for `crit.val`, `conf.int`, and/or `p.values`

[kylebutts]

I have been thinking about this proposal for a while in regards to the "fixest for developers" pitch. Here is my thinking:

• Say I do a bootstrap procedure to generate the vcov. I might want to use empirical bootstrapped confidence intervals.
• I want to use a multiplier score bootstrap to avoid problems with multiple hypothesis testing, I might want to return the vcov + critical values to use (e.g. Bonferroni correction)
• I want to use the t-stat for cluster and/or small samples.
• I have an IV regression and I want to report Anderson-Rubin confidence intervals
• I want to use randomization inference to compute p-values
• https://cran.r-project.org/web/packages/effClust/index.html custom t-stat effective number of clusters

The solution I was envisioning is to allow a set of attr be attached to the vcov and be plugged in to summary/etable/iplot/conf.int/etc. I know it would be a lofty change but I think it will prove useful for making fixest a "backend" for a ton of packages. I would probably create a lot of these, tbh.

What do we think? I can give a stab at a PR if y'all think it's a good idea

[kylebutts]

I think a lot of this will work with a modification to coeftable.fixest, so maybe less work than I thought originally. Will keep digging and update progress here