C++ refactoring: ak.cartesian, ak.argcartesian

[ioanaif]

No description provided.

[codecov]

Codecov Report

Merging #1317 (4e143f7) into main (5613d44) will decrease coverage by 1.81%.
The diff coverage is 47.08%.

❗ Current head 4e143f7 differs from pull request most recent head 9682cf0. Consider uploading reports for the commit 9682cf0 to get more accurate results

Impacted Files	Coverage Δ
src/awkward/_v2/_connect/cling.py	0.00% <0.00%> (ø)
src/awkward/_v2/contents/listoffsetarray.py	80.55% <ø> (+0.13%)	⬆️
src/awkward/_v2/operations/structure/ak_unzip.py	75.00% <ø> (ø)
...wkward/_v2/operations/structure/ak_argcartesian.py	70.58% <64.28%> (-4.42%)	⬇️
...ard/_v2/operations/structure/ak_argcombinations.py	76.92% <70.00%> (+1.92%)	⬆️
...wkward/_v2/operations/structure/ak_combinations.py	90.90% <87.50%> (+15.90%)	⬆️
...c/awkward/_v2/operations/structure/ak_cartesian.py	89.34% <89.07%> (+14.34%)	⬆️
src/awkward/_v2/_lookup.py	97.50% <97.50%> (ø)
src/awkward/_v2/__init__.py	100.00% <100.00%> (ø)
src/awkward/_v2/_connect/numba/arrayview.py	96.76% <100.00%> (+0.51%)	⬆️
... and 15 more

[jpivarski]

Everything here looks great, including the very complicated cartesian function.

I noticed that argcartesian is excluding negative axis; I don't see a reason for that, so I'm removing that restriction and adding a test. Once that passes, I'll merge this.

Thanks!

[jpivarski]

I see no reason why the axis can't be negative. It's like that in v1, also:

https://github.com/scikit-hep/awkward-1.0/blob/5e4e4b984a62baf4075ab96113903f49de058ada/src/awkward/operations/structure.py#L3598-L3602

I'm guessing it's an artifact from a very early time. Anyway, I just committed this fix and added a test, so this PR will relax that constraint.