🎨 workaround for path-dependent pyright bug by shuffling disjoint overloads

[jorenham]

When jorenham/optype@a0b79fc is installed, these changes will resolve the 75 false-positive pyright errors that are reported when running pyright tests or basedpyright tests (without a path these weren't reported).

The changes here should not directly affect the users of static type-checkers other than pyright (assuming they follow the typing spec correctly).

This workaround is a lot more subtle than #742, but requires a releasing a new optype version that removes the first type parameter default for optype.numpy.CanArrayND (which is a breaking change).

Personally, I prefer this workaround over #742. So I'll work on a new optype 0.12 release for this. It's no problem to merge this before that time, since this doesn't cause any new false-positives without the new optype release.

cc @JulVandenBroeck

[jorenham]

I've confirmed that together with jorenham/optype#354, this will indeed resolve all 75 false positives that are currently reported when running uv run basedpyright tests.

Note that the next optype release will include other features as well, so I won't be able to release it directly. I guess it'll be tomorrow or the day after.