Vector3: correct the randomDirection() method

[WestLangley]

Follow-on to #22494

three.js uses a 'y-up' convention. The current implementation was copied from a source using a 'z-up' convention. This PR fixes that oversight.

Note: the prior code did generate uniform samples, in spite of the oversight.

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

Filesize dev	Filesize PR	Diff
676.1 kB (168.2 kB)	676.1 kB (168.2 kB)	-4 B

🌳 Bundle size after tree-shaking

Minimal build including a renderer, camera, empty scene, and dependencies.

Filesize dev	Filesize PR	Diff
457.2 kB (111 kB)	457.1 kB (111 kB)	-4 B

[WestLangley]

The screenshots I generate are not correct for some reason.

[Mugen87]

I'll merge and try it on my end.

[LeviPesin]

I believe there is no difference in the behaviour of the previous and new code -- they both generate a uniform sample on the sphere, it wouldn't depend on using z-up or y-up (it is uniformly distributed in any case -- a rotation of uniform distribution is still the same uniform distribution).
The difference in the screenshots happens only because of changing the order of Math.random() calls.
@Mugen87 @WestLangley

[WestLangley]

Here is the screenshot of webgl_raycaster_bvh.jpg generated on my iMac:

npm run make-screenshot webgl_raycaster_bvh