fix(rust,python): Fix edge-case where the Array dtype could (internally) be considered numeric

[alexander-beedie]

Came across this one after seeing some inconsistent error messages debugging something else.
(Addresses an edge-case where Array dtype could inadvertantly be considered numeric).

• The is_numeric check was a bit peculiar (and likely prone to future breakage if/when adding new dtypes) as it decided that a dtype was numeric if It wasn't various other dtypes, rather than because it was numeric. Changed this so that we positively assert the dtype from the known numeric types instead, which seems a lot cleaner.

---

Also: fixes a minor test lint (mypy on py 3.11) caused by an update to pandas-stubs.

Note: for some reason we don't currently consider Decimal to be numeric here; that seems odd, so will probably revisit to better-understand why... this PR does not change this behaviour (indeed, quite a few things blow up if you do ;)

[stinodego]

I was actually staring at that is_numeric implementation yesterday... I'd say your implementation is much better 👌

for some reason we don't currently consider Decimal to be numeric

Ritchie explained this to me yesterday (I had also noticed this). It's not numeric because you cannot do regular arithmetic with it.

I'm always happy with more pytest.mark.parametrize 😄 I added a small commit to add the correct type hint and remove the warning filter (apparently it was unnecessary).

[alexander-beedie]

Ritchie explained this to me yesterday (I had also noticed this). It's not numeric because you cannot do regular arithmetic with it.

I figured it would be something like that - good to know :)
Maybe we can add a more inclusive numeric method for other uses later...

I'm always happy with more pytest.mark.parametrize 😄 I added a small commit to add the correct type hint and remove the warning filter (apparently it was unnecessary).

Ahh, well spotted!

[stinodego]

Maybe we can add a more inclusive numeric method for other uses later...

Yeah this begs the question 'what exactly is numeric' and maybe we should be calling it differently... not super simple to answer though. I decided to leave it alone for now.