compare --> loo_compare

[jgabry]

This PR deprecates the compare() function in favor of loo_compare(). This is for several reasons:

• compare() is not a generic with methods, which results in rstanarm, brms, etc., using different names for wrappers around compare(). With loo_compare() we can have methods loo_compare.stanreg, loo_compare.brmsfit, etc, unifying the names.

• compare() also conflicts with function names in other packages (e.g. testthat::compare)

We should time the release of this so that the loo_compare() methods in brms and rstanarm are released at the same time.

[paul-buerkner]

Hey Jonah! I will look at the code tomorrow and test it together with the loo_compare stuff of brms to make sure everything is compatible.

[paul-buerkner]

The code looks good to me and works smoothly with the corresponding changes I made in brms.
I think this is ready to be merged.

[paul-buerkner]

One little thing: Could we add an is.kfold function as well?

[jgabry]

I'm ready to merge this but @avehtari do you want to first try out the loo_compare() function (on the loo_compare) branch and let us know if you like the interface/output?

[codecov-io]

Codecov Report

Merging #93 into master will decrease coverage by 1.07%.
The diff coverage is 86.07%.

@@            Coverage Diff             @@
##           master      #93      +/-   ##
==========================================
- Coverage   94.31%   93.24%   -1.08%     
==========================================
  Files          15       17       +2     
  Lines        1197     1259      +62     
==========================================
+ Hits         1129     1174      +45     
- Misses         68       85      +17

Impacted Files	Coverage Δ
R/kfold-helpers.R	100% <ø> (ø)	⬆️
R/helpers.R	98.52% <ø> (-0.09%)	⬇️
R/print.R	92.45% <0%> (-5.55%)	⬇️
R/kfold-generic.R	0% <0%> (ø)
R/psis.R	90.66% <100%> (+0.06%)	⬆️
R/loo.R	92.61% <100%> (+0.07%)	⬆️
R/compare.R	83.63% <100%> (-11.75%)	⬇️
R/waic.R	100% <100%> (ø)	⬆️
R/loo_compare.R	91.04% <91.04%> (ø)
... and 2 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update db9fd3f...190f79a. Read the comment docs.

[jgabry]

@paul-buerkner I removed the deprecation warning from compare() for now as we discussed. I'm running into a few issues with reverse dependency checks and I need to sort out if they are loo related, but otherwise I think this is ready to go.