fix NA filled by shift(nanotime, fill=NA)

[mschubmehl]

Closes #3945

Currently shift(nanotime(1:4)) incorrectly fills a numeric NA rather than an integer64 NA into the result vector. This is because we incorrectly check that the class is exactly integer64, rather than inherited from it, as we do in most other contexts. This PR fixes that, and adds test cases for shift called on nanotimes.

[codecov]

Codecov Report

Merging #3942 into master will decrease coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master   #3942      +/-   ##
=========================================
- Coverage    99.4%   99.4%   -0.01%     
=========================================
  Files          72      72              
  Lines       13643   13642       -1     
=========================================
- Hits        13562   13561       -1     
  Misses         81      81

Impacted Files	Coverage Δ
src/shift.c	100% <100%> (ø)	⬆️

---

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 a004873...e449d6b. Read the comment docs.

[MichaelChirico]

LGTM. Could file an issue as well that we could close with this? Easier to find it there if anyone else notices the issue before next release

[jangorecki]

LGTM

[mschubmehl]

Closes #3945, and NEWS.md updated to reflect that. Thanks, all!