Replacing Shader Code for FMA

[Malidos]

I added fma into all shader snippets I could identify, fulfilling #26.2, but since fma only accepts variables of the same type I had to replace some float constants with vectors.

fma combines multiplication and addition into a single operator which should double the speed of that specific operation, but since the Shader contains many other more complex operations, this resulted in an insignificant performance increase of about 1% which makes me doubt whether this should be merged at all, but that's up to you.

It should be compatible with the changes in #500 because it introduces a replacement macro for fma.

[TokisanGames]

Thank you. When making PRs you need to make a branch in your repository, then make a (Pull) Request to merge your branch into our main. Never your main to my main, and never add merge commits into your PRs. Never use git merge, nor "sync" on gh desktop. You can probably fix it by dropping the merge commit; I can probably as well.

Read and follow the Godot PR Workflow (and all our contributor docs).

At the least, I won't merge until 500 is merged, then we can rebase this and look at it then. I'll want to try it on my system and get more data points.

Why the changes from vec4s to mat4s?

@Xtarsia Thoughts?

[Xtarsia]

any case where there are extra steps needed just to make fma() work, are usually not worth it. the same applies to nesting fma() instructions, which can prevent the compiler from being able to parallelize something it might have before.

the swap to mat4 and the following nested fma() are likley not gaining anything - and may even be slowing things down. But other areas like the rotation functions could well be faster.

it is handy to have this laid out with all currently possible places where fma() could be put.

[TokisanGames]

@Xtarsia Ok, I'll go through and take the simpler ones, especially in utility functions, and skip the nested fmas and conversions to mat4s and vec3s before release.