Fix comparisons between Series and cudf.NA

[brandon-b-miller]

Fixes #7043, gives less than ideal results due to #7066.

[codecov]

Codecov Report

Merging #7072 (1550ba9) into branch-0.18 (8860baf) will increase coverage by 0.03%.
The diff coverage is 100.00%.

@@               Coverage Diff               @@
##           branch-0.18    #7072      +/-   ##
===============================================
+ Coverage        82.09%   82.12%   +0.03%     
===============================================
  Files               97       97              
  Lines            16474    16498      +24     
===============================================
+ Hits             13524    13549      +25     
+ Misses            2950     2949       -1     

Impacted Files	Coverage Δ
python/cudf/cudf/core/column/string.py	86.65% <100.00%> (ø)
python/cudf/cudf/core/series.py	91.10% <100.00%> (-0.06%)	⬇️
python/cudf/cudf/core/column/numerical.py	94.17% <0.00%> (-0.24%)	⬇️
python/cudf/cudf/core/column/lists.py	91.66% <0.00%> (-0.09%)	⬇️
python/cudf/cudf/core/reshape.py	91.00% <0.00%> (-0.04%)	⬇️
python/cudf/cudf/_fuzz_testing/fuzzer.py	0.00% <0.00%> (ø)
python/cudf/cudf/utils/hash_vocab_utils.py	100.00% <0.00%> (ø)
python/cudf/cudf/core/dataframe.py	90.76% <0.00%> (+0.05%)	⬆️
python/cudf/cudf/core/dtypes.py	90.50% <0.00%> (+0.12%)	⬆️
... and 3 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 8860baf...1550ba9. Read the comment docs.

[brandon-b-miller]

rerun tests

[galipremsagar]

Should we add some unit-tests so that we can be sure we don't regress on this fix in the future while changing this part of code?

[brandon-b-miller]

@galipremsagar yes we should add tests. When I sat down to write them though, the issue of #7066 came up. This PR causes the code to "work", but we get nan-like results. I am not sure if I should write tests that assert that logic since it's arguably not right.

[galipremsagar]

Got it, probably we can just have an eq test case and xfail rest of the comparisons(lt, gt, ...) mentioning the #7066 in xfail reason.