PhysicsDirectBodyState3D: report contact points using global coordinates

[rburing]

Also update the documentation to reflect this in both 2D and 3D. (It was already the case in 2D, just not documented.)

Follow-up to #58880

• Fixes PhysicsDirectBodyState2D: get_contact_local_position and get_contact_collider_position are inconsistent with 3D analogues #76050

This is the most pragmatic solution, keeping 2D as it has always been (since Godot was made open source), and updating 3D which was bugged for a long time, then fixed recently by #58880 to be self-consistent but inconsistent with 2D; now this PR makes it consistent with 2D.

[rburing]

Technically breaks compatibility, but only in methods which were just recently made bug-free.

[adamscott]

I agree with the change. More consistency is desired between 2D and 3D features.

[clayjohn]

Adding offset_a to global_b seems a little bit fishy to me. But I'm pretty unfamiliar with this code. Could you explain why offset_a is used and not offset_b as in the previous code?

[rburing]

@clayjohn The reason is basically this:

godot/servers/physics_2d/godot_body_pair_2d.cpp

Line 266 in a7276f1

//use local A coordinates to avoid numerical issues on collision detection

The global_A and global_B vectors are the coordinates of the contact points in a coordinate system which is centered on body A (and has the same orientation as the global coordinate system). So to translate to the global coordinate system you add offset_A, which is defined as body A's origin in global coordinates.

Edit: For completeness, the previous code was reporting global_A and global_B - offset_B, where offset_B = B->get_transform().get_origin() - A->get_transform().get_origin(), so it was reporting offsets from body origins, as noted in the linked issue that this PR closes.

[akien-mga]

Thanks!