Fixes for KP stat calculation

[karldw]

Fixes #607

• This PR fixes the error Error in solve(t(Gmat)) %*% PI by ensuring PI is a matrix rather than a dataframe (as mentioned in Kleibergen-Paap test code broken; Effective F-statistic implementation #161 (comment))
• It adds a warning for the case with clustering and n_endo ≠ n_inst, but doesn't implement the functionality. The code now returns an NA value for that case.
• It adds a couple of tests that run these cases and updates the URL for https://github.com/FixedEffects/Vcov.jl

[lrberge]

Thanks @karldw!

One minor things: Could you be more precise on the warning message by telling when the KP stat is actually implemented?

[karldw]

Thanks! Does this edit fit what you had in mind? I'm happy to make any changes.

[lrberge]

Sorry for the late reply. And sorry for yet other comments:

• can you return(NA) directly right at the beginning of the function (circa line 1626)? And not wait for the case to pop later on (this avoids unnecessary calculations)?
• can you drop the last line in the warning (this is implicit)?

I'm also happy to make the changes when I get the chance in January.

[lrberge]

All good, thanks. Sorry for the back and forth.

[karldw]

No problem at all -- thanks for reviewing!