implement per-stage normalscale

[illwieckz]

This PR sits above #230.
This is a demonstration of what ca be achieved thanks to #230.

Previously in #180 I introduced a per-material normalFormat option that would be able to flip channels. But then I discovered that other engines relies on a normalScale option that can both scale (and flip) channels.

A proper implementation required to be done on the stage scope.

Note that ioq3 seems to only read normalScale X Y while we are able to read normalScale X Y Z, if Z is missing, default one is used instead, to keep compatibility with material that may already exists.

As an example, we can look at some materials from tex-vega_src.dpkdir that have a flipped X channel.

Before:

After:

They can be easily compared this way:

or with a slider there: https://imgsli.com/OTI4Mw

To do fix them, their materials were edited from this:

textures/shared_vega/panel01
{
	qer_editorImage textures/shared_vega_src/panel01_d

	diffuseMap  textures/shared_vega_src/panel01_d
	normalMap   textures/shared_vega_src/panel01_n
	specularMap textures/shared_vega_src/panel01_s
}

textures/shared_vega/trim03
{
	qer_editorImage textures/shared_vega_src/trim03_d

	diffuseMap  textures/shared_vega_src/trim03_d
	normalMap   textures/shared_vega_src/trim03_n
	specularmap textures/shared_vega_src/trim03_s
}

to this:

textures/shared_vega/panel01
{
	qer_editorImage textures/shared_vega_src/panel01_d

	{
		diffuseMap  textures/shared_vega_src/panel01_d
		normalMap   textures/shared_vega_src/panel01_n
		normalScale -1 1 1
		specularMap textures/shared_vega_src/panel01_s
	}
}

textures/shared_vega/trim03
{
	qer_editorImage textures/shared_vega_src/trim03_d

	{
		diffuseMap  textures/shared_vega_src/trim03_d
		normalMap   textures/shared_vega_src/trim03_n
		normalScale -1 1 1
		specularmap textures/shared_vega_src/trim03_s
	}
}

In the process we also moved from separate stages that has to be collapsed by the engine to a multi texture stage that is collapsed from the start.

---

Also note we inherited from XreaL another normalScale keyword that is currently only used by broken liquid shader, and this keyword is expecting an expression. We have to rename it. It is now named normalIntensity, this applies on XY normal map channels and can be computed on every frame by using expressions.

It would have been better to keep the normalScale keyword for that feature (as exists fresnelScale for example) and use normalChannelScale for the par-channel-scaling-and-flipping variable, but normalScale is already widely used by other idtech3 derivated engines, while our liquid shader is used by no one yet, even not by us.

[illwieckz]

This seems ready to me.

[illwieckz]

I added a commit to remove the r_normalScale glsl define. Since we pass u_NormalScale and compute it in all case, it may be better to feed u_NormalScale with r_normalScale value if exists.

[illwieckz]

I noticed I forgot to make the dead liquid shader using u_normalScale, that's now fixed.
This is just to not let dead code rot until we revive it.

[illwieckz]

@slipher, I added a small commit I want your advice on.

This commit is just done to reduce boilerplate. But it replaces a definition by an affectation, do you think a compiler is able to optimize it, or, if not, this is negligible?

If you think it's OK I'll squash it, otherwise I'll drop it.

[slipher]

Removing r_normalScale define seems good. Not sure what you mean about stuff needing to be optimized as you simply removed the extra instructions from GLSL.

[illwieckz]

@slipher: Not sure what you mean about stuff needing to be optimized

instead of this in 8 places:

		vec3_t normalScale = { 1, 1, 1 };
		SetNormalScale( pStage, normalScale );

I do this in 8 places:

		vec3_t normalScale;
		SetNormalScale( pStage, normalScale );

And this in 1 place (in SetNormalScale):

 		normalScale[ 0 ] = 1;
		normalScale[ 1 ] = 1;
		normalScale[ 2 ] = 1;

The code is more compact this way, and less prone to errors, but in first case the vec3 is defined to 1,1,1, in the second case the vec3 is defined to 0,0,0 then set to 1,1,1.

[illwieckz]

Does anyone have something to say against that?