Fix iridescence energy conservation for mixed materials

[mrdoob]

Related issue: #31350

Description

This PR fixes an issue where iridescence was not correctly calculated for materials with intermediate metalness values (mixed dielectric/metallic), leading to incorrect energy conservation.

Before	After

Changes

• Split iridescenceFresnel into iridescenceFresnelDielectric and iridescenceFresnelMetallic in the PhysicalMaterial struct.
• Updated lights_fragment_begin.glsl.js to calculate these separately using specularColor (0.04) and diffuseColor (base color) respectively.
• Updated lights_physical_pars_fragment.glsl.js to use the correct Fresnel term for dielectric and metallic paths in RE_IndirectSpecular_Physical.
• Applied the same logic to WebGPURenderer in src/nodes/functions/PhysicalLightingModel.js, splitting iridescenceFresnel and iridescenceF0 calculation and usage.

This aligns the behavior with Khronos glTF expectations and fixes the issue reported in #31350.

(Solved in a single prompt using https://antigravity.google/ with Gemini 3.0)

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

	Before	After	Diff
WebGL	356.28 86.49	356.68 86.53	+391 B +43 B
WebGPU	616.52 173.1	616.83 173.15	+309 B +43 B
WebGPU Nodes	615.13 172.85	615.43 172.9	+309 B +44 B

🌳 Bundle size after tree-shaking

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

	Before	After	Diff
WebGL	488.15 121.31	488.54 121.36	+391 B +42 B
WebGPU	688.76 189.02	689.07 189.06	+309 B +42 B
WebGPU Nodes	630.48 172.25	630.48 172.25	+0 B +0 B

[mrdoob]

@sunag Does the TSL port look good to you?

[sunag]

@sunag Does the TSL port look good to you?

Yes, it looks great to me.

[sunag]

(Solved in a single prompt using https://antigravity.google/ with Gemini 3.0)

👀