Refine logic to determine internal texture format for shadow pass

It seems as if 32-bit float textures are not considered filterable in WebGL 2 [1].
The necessary extension for those textures does not change that, either [2].
With this in mind, we should redesign the algorithm to determine bitness, number of channels and specific internal format, and filtering method for the shadow map of a shadow pass.

Possible combinations seems to be:
* RED with R8 and linear filtering
* RED with R8 and nearest filtering
* RG with RG8 and linear filtering
* RG with RG8 and nearest filtering
* RGBA with RGBA8 and nearest filtering

Sources:
[1] https://developer.mozilla.org/en-US/docs/Web/API/WebGLRenderingContext/texImage2D)
[2] https://developer.mozilla.org/en-US/docs/Web/API/OES_texture_float