Initial implementation of specular haze

[jstone-lucasfilm]

This changelist adds an initial implementation of specular haze in the reference MaterialX implementation of OpenPBR Surface, as specified in #254.

[portsmouth]

Review: reference/open_pbr_surface.mtlx

The implementation correctly encodes the spec. Notes by section:

I checked the MaterialX file compiles in the validator.

Haze roughness computation

• haze_roughness_base = mix(bg=specular_roughness, fg=1.0, mix=spread) → lerp(r_core, 1, ξ_h) ✅
• Coat roughening mirrors the existing core treatment: (min(1, 2·r_coat⁴ + r_haze⁴))^0.25 blended by coat_weight ✅
• Anisotropy applied via the same open_pbr_anisotropy macro with specular_roughness_anisotropy_clamped, matching the spec ("anisotropy of the haze lobe is identical to that of the core") ✅

Dielectric haze

• dielectric_reflection_haze / _haze_tf faithfully mirror the core dielectric_reflection / _tf nodes with main_roughness replaced by haze_roughness; all other inputs (tint, ior, scatter_mode, normal, tangent, thin film) are identical ✅
• dielectric_reflection_with_haze = mix(fg=haze_tf_mix, bg=core_tf_mix, mix=specular_haze) → (1 − w)·f_core + w·f_haze, matching the spec formula ✅

Metal haze

• metal_bsdf_haze / _haze_tf mirror metal_bsdf / _tf with haze_roughness substituted; all parameters (weight, color0, color82, normal, tangent, thin film) copied correctly ✅
• metal_bsdf_with_haze uses the same mix ordering ✅

Clamping & wiring

• Both specular_haze and specular_haze_spread have clamp nodes placed consistently alongside the other clamped inputs ✅
• dielectric_base.top → dielectric_reflection_with_haze, base_substrate.fg → metal_bsdf_with_haze ✅

No issues found.

[portsmouth]

LGTM

[AdrienHerubel]

LGTM