Add support for GL_EXT_texture_offset_non_const

[rg3igalia]

Closes #3765

[rg3igalia]

I noticed the constness of the offset argument was not being checked with sparse texture functions, so I've added a commit on top with some additional code and tests.

[arcady-lunarg]

Should these test cases be generating SPIR-V? I feel like that might be useful here, since ultimately this is an extension for use with Vulkan and SPIR-V.

[rg3igalia]

Should these test cases be generating SPIR-V? I feel like that might be useful here, since ultimately this is an extension for use with Vulkan and SPIR-V.

Yes, it makes total sense. I wasn't aware of the distinction between the different test groups as I'm not familiar with glslang's code base. Thanks for the tip!

[rg3igalia]

I see the CI failures are related to the shaders failing validation and suggesting something like this:

--- Test/baseResults/validation_fails.txt	2024-11-05 14:19:47.844071917 +0000
+++ -	2024-11-05 14:25:25.835244652 +0000
@@ -68,4 +68,6 @@
 Test/baseResults/spv.separate.frag.out
 Test/baseResults/spv.sparseTextureClamp.frag.out
 Test/baseResults/spv.sparseTexture.frag.out
+Test/baseResults/spv.sparsetextureoffset_non_const.vert.out
+Test/baseResults/spv.textureoffset_non_const.vert.out
 Test/baseResults/vk.relaxed.errorcheck.vert.out

Which may be solved once the SPIR-V Tools issue is solved. So what should I do in the mean time? Update the file in this MR?

[arcady-lunarg]

We can wait until the SPIRV-Tools change is merged and the you can update the known_good.json to point to the new SPIRV-Tools commit.

[rg3igalia]

Alright, feel free to merge if you want in the mean time. The CTS tests using this are still under review so I'm not in a hurry.