Expose some useful method from rotation class Quaternion & Basis & Transform2D to GDScript

[TokageItLab]

Expose useful methods to GDScript for calculating rotations and scales.

[TokageItLab]

Added some functions to Transform2D for scale, but I am not sure if the behavior is correct, so I will test it later.

[TokageItLab]

Finished testing the methods added in Transform2D. The calculation seems to be correct.

[reduz]

The problem with a method like this is that its very misleading. You cant call multiple times without incurring into precision loss. The size will mutate over time and deform the object.

[TokageItLab]

The same thing has already been done in Basis, but I can't think of any other way. Definitely there is an accuracy issue, but I assume that calling it multiple times (consectively) seems to be a corner case. Should I describe the possible loss of accuracy in the documentation?

[TokageItLab]

Added notes:

Note: When a Basis/Transform2D that is already orthogonalized uses this method, the equality of the values before and after is not guaranteed.

[TokageItLab]

Instead of adding notes, I added a check is_orthogonal() to Basis and added a check Math::is_zero_approx(get_skew()) to Transform2D for making guarantee the equality.

[reduz]

One possibility for the Transform2D version is to always make the Y axis orthogonal to the X axis and leave the X axis untouched, this should more or less avoid precision errors.

[TokageItLab]

I implemented what you said, but the behavior is no different from setting skew = 0, so I don't think there is much point in having it. Well, I think it is fine to have the function for consistency with 3D.

[TokageItLab]

for now

Vector2 y = Vector2(-(columns[0].y), columns[0].x).normalized(); // Make a vector orthogonal to x.
if (y.dot(columns[1].normalized()) < 0) {
	y = -y; // Flip to be closer to current y.
}
columns[1] = y * columns[1].length();

[AThousandShips]

Note that adding mutating methods to the binds is not possible and doesn't work, see #62706