[Core] Fix AABB.encloses failing on shared upper bound

[AThousandShips]

This differs from Rect2(i) and was fixed for those classes in the past

See:

• Make Rect2.encloses return true on same size #32478
• Fix incorrect Rect2i calculations: intersects and encloses #57469

Which fixed this for those types, makes sense to me to have parity between the two, and it doesn't make sense to only allow it on one side, or fail to enclose yourself. I'd also say that the following should be equivalent: a.encloses(b) and a.intersection(b) == b

Essentially: Either it should be true only when no sides touch, or it should allow any side to touch, an alternative would be to add an encloses_inclusive or some other wording that does what this does, and change all the checks for the encloses to be strict

(Will open a godot-cpp equivalent if this gets approved)

[aaronfranke]

Yup, this is good for consistency and I'd say matches user expectations. Approved.

[akien-mga]

Would like @lawnjelly's opinion on this, I remember a lot of bikeshedding last time we attempted changing such conventions.

[Sauermann]

The bikeshedding in #57469 was mostly about expand_to with the result, that it works as intended.
My understanding was, that the change to a1.encloses(a1) == true was the consensus. So this PR simply fixes a bug in my opinion.

[lawnjelly]

Yes this is fine imo. The difficult version is the integer bound where you can have exclusive or inclusive definition at upper border. 👍

[AThousandShips]

Will add some extra checks like these to the Rect2(i) in a separate PR to ensure compliance

And I'll open a stripped down 3.x one without the tests

[akien-mga]

Thanks!

[AThousandShips]

Thank you!

[YuriSizov]

Cherry-picked for 4.2.2.