Various export fixes for subsurface scattering

[MoritzBrueckner]

• The "Auto" option for SSS could lead to a missing sssWidth variable in the shader which is one of the issue mentioned here: Just few bug pictures #2136
• If multiple SSS nodes were present in a material but the first to be parsed was not used, the shader didn't use SSS. It's still far away from Blender's subsurface scattering, but it's a bit more reliable now.
• When compiling with SSS disabled and then setting it to "Auto" without clearing the cache, it could happen that the generated materials didn't activate SSS because the _SSS flag wasn't set. I'm not completely happy with the solution, but I'm not yet confident enough with the material exporter code to be able to refactor the code in a way that prevents the dependency/order issues that can happen. It might be a good idea to first check all the requirements of a material, store them temporary with the material and only then create the shaders.

@luboslenco Do you remember why the material ID is handled via a uniform whose value is defined during export (see this and this)? Is it for context variants or to be able to merge similar shaders for less context switches?

[luboslenco]

Thanks!

Is it for context variants or to be able to merge similar shaders for less context switches?

Yeah looks like it was done for less context switches.