Fix sqrt application for radius in particle ring emission

[Arnklit]

This is a follow up to #91223.

I did not apply the sqrt correctly in that PR. The radius control is broken.

I also failed to realise that the even distribution for a circle shape becomes more and more like a cylindrical strip the closer the inner radius get's to the outer radius. I think the best solution for this is to lerp between the two distributions.

The inner radius could also be above the the outer radius before, it is now clamped by the outer radius to avoid unexpected behaviour.

godot.windows.editor.dev.x86_64_JydpEjwWOD.mp4

Note if this not merged before 4.3, then #91223 has to be reverted.

[Arnklit]

OK, fixed as @kleonc suggested. Thanks for that. I think I panicked a bit with me making the regression in the first place and couldn't figure out the right way to resolve it.
I left the ability to have the inner radius go above the outer as it just starts acting as the outer radius now, which seems ok.

[kleonc]

@Arnklit Note you haven't updated cpu_particles_3d.cpp in your new commit (or rather there's a leftover in there). 🙃

[Arnklit]

@kleonc just shoot me. Fixed now.

[kleonc]

Haven't tested but calculations look good. 👍

[akien-mga]

Thanks!