Revert attempted fix of trimesh CCD

[rburing]

This reverts a part of #72917 which was not the correct solution to the problem.

• Fixes Roughly Perpendicular Collisions with Concave Polygon are not detected and stick #74756

The issue #72889 remains fixed.

Adding gravity_scale = 5.0 to the MRP from #72889 still shows a tunneling issue, hence the FIXME comment.

[fabriceci]

I'm looking at it tomorrow. Can you add tests to the test project, to reduce the risk of regressions?

[rburing]

@fabriceci I've added a test here:

• Add tests for rigid body CCD against trimesh/concave wall fabriceci/Godot-Physics-Tests#17

Regression test output with this PR:

This PR reverts an incorrect attempt to fix rigid body CCD against trimesh static bodies. That previous attempt did prevent tunneling (hence reverting it causes a regression), but it had the side effect of no longer detecting collisions, which is worse, since that issue can affect many projects, see #74756 (comment).

The test project uses the extreme speed 25000. As mentioned above, the regression does not happen for all bodies/speeds, e.g. #72889 remains fixed. In the test project the threshold is at speed 550.

[fabriceci]

LGTM

[YuriSizov]

Thanks!

[YuriSizov]

Cherry-picked for 4.0.2.