Clamp metallic Fresnel after `specular_weight` scaling via F0 and F82

As outlined in the previous meeting, this proposal by @fpsunflower to apply clamping (after `specular_weight` scaling) via the F0 and F82 colors was considered better than the current approach in v1.2:

>The team discussed implementing a scaled-back version of a previous change to address development contributions. They agreed to modify the implementation to clamp only the F0 parameter rather than scaling the entire Fresnel curve, as this approach would be more physically motivated and visually desirable. The group also identified an issue with layering R and T lobes in MaterialX, where the current implementation produces incorrect results when layering over identical parameters. They decided to create a GitHub issue to address this broader problem across different renderers and models, as it affects multiple systems including MaterialX and Llama.

The issue with the current approach is that it isn't doable in the MaterialX graph, since it requires plumbing into the Fresnel implementation in the BSDF. This proposal would fix that.

The issue names no files, tests, or entry points. Start by reviewing the current Fresnel clamping approach and the MaterialX graph constraint described in the issue. Done means agreeing on an implementation that clamps after specular_weight scaling via F0 and F82, while addressing the stated R/T layering concern.

Written by the indexing model from the issue text.