Daemon Make feature disablement more consistent and move some compute at stage configuration

The PR was split, please review those PR before reviewing this one:

Once those other PRs are merged, the remaining commits to review will be:

Original message:

Make feature disablement more consistent: various combinations of disabled r_precomputedLight where wrong.
Make possible to load deluxe maps even if light mapping is disabled, as deluxe mapping with grid lighting is possible.
Do not load lightgrid if precomputed lighting is disabled, the same way lightmap loading is already disabled.
Call Render_lightMapping() on every related shader stages because that function knows how to render in all cases.
Render_generic() calls Render_generic2D() if 2D, This allows to move the function from Tess_StageIteratorGeneric() which means many tests from Tess_StageIteratorGeneric() can be moved to shader stage configuration.
Select shader stage render function only once. Introduce SetShaderStageRenderers() and call it at shader stage configuration. Not only factorise the code, but makes sure the Render function as selected once and only once, if possible when finalizing the shader stage, instead of running it on every frame
Set colorGen and alphaGen only once. Factorise the code, do it only once per shader stage, and do it when finalizing the shader stage if possible.
Make r_precomputedLighting a cheat cvar. Disabling this cvar disables all precomputed shadows and then makes everything fullbright, which gives an unbalanced advantage by making everything visible.

Mar 30 '23 22:03 illwieckz

The two biggest points would be summarized that way:

Disabling r_precomputedLighting now works correctly, and it is now a cheat cvar (disables all precomputed lights and then shadows).
Some code to select the Render_* function and sets some colorGen and alphaGen options are now run only once for the whole game time instead of doing the related computations on every frame. This unifies many Render_* function signatures to be the same. It means that on every frame for every shader stage the code now just runs the related function and that's all. The code that is optimized wasn't super heavy on CPU (it was mostly tests) but that design bothered me for a long time, and the rewrite also brings some factorisation and code deduplication, also reducing line count. Less code to read for humans, less code to execute for CPUs.

Mar 30 '23 22:03 illwieckz

I'm splitting the PR.

Mar 31 '23 08:03 illwieckz

Please review those PR before reviewing this one:

Mar 31 '23 08:03 illwieckz