MeshPhysicalMaterial: Fix Anisotropic regression.

[Mugen87]

Fixed #33201.

Description

The PR corrects the changes from #32330. Removing the saturate() was right but it was missed to introduce the same guard as V_GGX_SmithCorrelated() to prevent division through 0. That produces NaN values and the reported black pixels.

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

	Before	After	Diff
WebGL	359.97 85.54	359.97 85.53	-6 B -8 B
WebGPU	632.12 175.39	632.13 175.39	+8 B +0 B
WebGPU Nodes	630.24 175.1	630.25 175.1	+8 B +0 B

🌳 Bundle size after tree-shaking

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

	Before	After	Diff
WebGL	492.17 120.12	492.17 120.11	-6 B -8 B
WebGPU	704.55 190.19	704.56 190.19	+8 B +2 B
WebGPU Nodes	653.78 177.6	653.78 177.6	+0 B +0 B