Standard Surface: Connect thin_walled to the surface constructor

[msuzuki-nvidia]

Connect thin_walled to the surface constructor in standard_surface nodedef.

[jstone-lucasfilm]

Thanks for putting this together, @msuzuki-nvidia. The change looks right on both counts: forwarding thin_walled into the surface constructor, and marking the interface input uniform="true" to match it.

@ld-kerley I'm CC'ing you here since you raised the backwards-compatibility question about changing uniform on a shipped interface input, which is a fair point to weigh. A few concrete things make me think the risk here is quite low:

• Established precedent: OpenPBR made the identical change (forward plus uniform="true") in Connect geometry_thin_walled to surface in OpenPBR #2759, so this brings standard_surface in line with a convention that has already shipped.
• It is the norm rather than a new pattern. The booleans that gate surface and BSDF construction are already uniform, including energy_compensation and retroreflective on the BSDF constructors, and the ND_surface constructor's own thin_walled. The non-uniform thin_walled interface on standard_surface is currently the outlier.
• Required for consistency once connected: the ND_surface input is uniform="true" and the MDL implementation declares it as uniform bool, so a non-uniform interface value feeding it would be a qualifier mismatch.
• No content regresses. Across all bundled materials and our render test suite, thin_walled is only ever authored as a literal value, and is never connected to an upstream node. The only thing affected would be a third-party document driving it from a varying source, which is not a meaningful operation for a structural surface flag of this kind.

Given that, I believe this is the right next step, but I'd like to confirm this with @ld-kerley before proceeding. If we are aligned, I would suggest we also note the uniform change in our changelog and release notes for discoverability.