Fix AABB calculation in Mesh::take_gpu_data

[oliver-dew]

Objective

• Fixes 0.18.0-rc.1 Aabb3d::new is being passed the wrong arguments #22169

Solution

• Mesh::take_gpu_data was calling Aabb3d::new and passing it min and max, but the constructor expected center and half size. This caused a panic for meshes where the vertices are all negative on any of the axes. And also means the aabb is incorrect, maybe causing issues like incorrect frustum culling? After a bit of discussion on Discord, the new constructor expecting center & half extents was deemed footgunny. So this PR renames Aabb3d::new to Aabb::from_half_size, and adds a new constructor Aabb::from_min_max. Mesh::take_gpu_data now calls from_min_max, fixing the AABB calculation bug.

Testing

• I hit this bug with a model where all the vertices were negative in one dimension (so the max is all negative, and when it was wrongly passed as the half size, fails the assertion that half size is >= 0). Confirmed that this change fixes the issue.
• This PR adds a unit test mimicking the above scenario, that panics without the fix
• How can other people (reviewers) test your changes? They could test it with a model whose max is less than zero in one dimension, or run the unit test
• Tested on macOS

[github-actions]

Welcome, new contributor!

Please make sure you've read our contributing guide and we look forward to reviewing your pull request shortly ✨

[oliver-dew]

Closed in favour of #22180