ShaderProcessor partially updating shader source to support uniform buffers

[mvaligursky]

a progress towards #4261

• a GLSL shader processing in order to allow WebGL shader to be transformed info a format we can transpile to WebGPU. Initially in this PR, attributes, varyings and uniforms are extracted, and transforms into required form - which means a fixed slot allocation, as well as uniform buffers.
• ProgramLibrary has been updated to allow this processing to be executed globally on any shader, which includes shaders set by the user directly on the Material, in addition to Standard, Basic, Sky, and Particle.

Additional changes:

• small clean up to material's handling of transparency functions.

Note: The shader processing does not currently execute on WebGl, but some shader functionality refactoring is used.

[slimbuck]

It seems like we're adding an invisible layer of processing to our shader code that will ultimately be difficult for users to debug if and when it goes wrong. I'm a little concerned that we're not really documenting the expectations or limitations of this step and it will ultimately become a magical black-box.

Are you happy with our approach of regex'ing shaders like this and processing the code in this way? Or do you expect we'll be rewrite this using different tools or ...?

[GSterbrant]

Approved with a few comment/formatting comments.

[mvaligursky]

Are you happy with our approach of regex'ing shaders like this and processing the code in this way? Or do you expect we'll be rewrite this using different tools or ...?

It's hard to know at this stage. Right now my focus is to transpile GLSL shaders so that I can directly execute them on WebGPU device, allowing me to develop the rest of the WebGPU rendering. This is a black box, but a pretty small one, and executes for now only on WebGPU.

There is also an option to execute this on WebGl to add support for Uniform Buffers in the future.

Even if we went ahead and rewrote all shaders in the WebGPU native shader language, due to the current lack of support of reflection for those, we'd need to do processing there as well and manually assign slots.