RFC: Move ScatterplotLayer to UBOs

[felixpalmer]

Follow up to #8782, trying out moving ScatterplotLayer to use UBO.

Change List

• Add ShaderModule for binding scatterplot uniforms
• Update shaders
• Set shaderInputs rather than uniforms in draw()

[coveralls]

coverage: 89.826% (+0.005%) from 89.821%
when pulling ed9343e on felix/layer-ubo
into ff9246e on master.

[ibgreen]

Nit: avoid default exports in new code?

[ibgreen]

Not something we need to sort now:

• I get the concern about typing twice and getting errors, but this extra module and file for just uniforms is just too much plumbing in my opinion that fragments code and adds as much inefficiency as it solves

There are other possible approaches we should consider,

• e.g. autogenerating the blocks from the shader module props.
• adding a common, or uniforms field to shader modules,
• etc.

[felixpalmer]

I would prefer to agree on a pattern that everyone is on board with, to avoid doing this work twice. Could you elaborate on how you'd expect the auto-generation to work? Are you saying we take the value of uniformTypes and produce the uniformBlock?

[ibgreen]

We should standardize on Settings vs Props for uniform inputs. I have gravitated towards ...Props in recent shadertools refactorings, but the question could still be considered open.

[felixpalmer]

Given how close it is to the ScatterplotLayerProps, it could be cleaner to just have the ShaderModule accept those and do the conversion in getUniforms

[ibgreen]

Thought: Maybe should be part of the layer test utils at some point?