MeshStandardMaterial: Remove roughness-dependent Fresnel from environment lighting

[WestLangley]

Correction to #15644 and #16559.

The sample code in the paper is not consistent with the derivation in the paper.

The derivation uses specular reflectance F0, the code used fresnel F.

This discrepancy has been noted elsewhere.

[WestLangley]

Even with this change, CI passed for me. We'll see what happens... :-)

[mrdoob]

Curious to see! 👀

[mrdoob]

Thanks!

[mrdoob]

It's pretty subtle in general, but I can see the change in the helmet:

Before	After

@elalish this will probably affect https://modelviewer.dev/fidelity/

[mrdoob]

More noticeable in the FlightHelmet with Model Viewer's lighting setup:

Before	After

[mrdoob]

I think this should bring us closer to the other renderers 👏

Notice how bright the surface of the wood piece was before this change:

Edit: After this change...

[mrdoob]

Also noticeable in Cerberus.obj:

Before	After

[mrdoob]

One more:

Before	After

[mrdoob]

It's surprising that the e2e ci hasn't noticed any change... 😶

[mrdoob]

As in #16559 (comment), it's specially noticeable in the tubes behind the helmet:

Before	After

[mrdoob]

Looking good! ✨

[elalish]

Thanks for the diligence, @WestLangley! I'm working on updating our fidelity tests, so I'll be able to give better feedback soon, but I agree with @mrdoob's analysis.