test() distinguishes NA, NaN in vectors

[MichaelChirico]

NB: targeting #7863

Closes #7853

1. test(0, data.table(a=NA_real_), data.table(a=NaN)) currently passes. I'll mark that for follow-up since, IINM, it will require a breaking change to all.equal.data.table().
2. froll.Rraw is not currently passing. @jangorecki could you PTAL and see if we should just update the tests or if this is exposing some unintended behavior?

[MichaelChirico]

Testing out the new stacking feature on GitHub. Locally, I'm still using jj to manage chains.

Let's see how it works! Ideally they work fluidly together.

[github-actions]

• HEAD=test-na-nan much slower for transform improved in #5493
  

Generated via commit aed009e

Download link for the artifact containing the test results: ↓ atime-results.zip

Task	Duration
R setup and installing dependencies	3 minutes and 13 seconds
Installing different package versions	48 seconds
Running and plotting the test cases	5 minutes and 35 seconds

[jangorecki]

From R-devel ultimate conclusion was to not depend on "NA vs NaN" -ness as it is platform dependent,

*) As Ista Zahn said (there, on R-devel): See ?NA, which says

"Numerical computations using ‘NA’ will normally result in ‘NA’: a
possible exception is where ‘NaN’ is also involved, in which case
either might result"

and as seen in this link, compiler flag dependent too.
#3353

we could possibly add one more section to froll manual explaining that matter.