Move more graphics operations to the GPU and shaders

[ultraq]

I've got a lot of image processing methods which I think can be moved to shaders in the GPU, instead of the Groovy code run on the CPU. These include:

• [ ] applying palettes to frames of animations/videos (the applyPalette ByteBuffer extension). I have a palette shader that can do this now, so it might be good to get the animation/video ones using it too Have moved to preloading textures data as RGB(A), so no palette shader any more
• [x] scaling images up to include more pixels so that linear texture filtering produces sharper results (the scale ByteBuffer extension, used via the --scale-low-res CLI flag)
• [x] scanlines! (is currently being done on the CPU as a fake texture)

Trying out the first item, something I noticed was that you can't reliably use GL_LINEAR on paletted images because then it'd interpolate the palette indices which doesn't guarantee an in-between colour. So being able to create a filter in the fragment shader as well will be good, and then I can enable it on a per-material basis instead of globally via the current CLI flag (--filter).

Some prior art/resources:

• https://github.com/rsn8887/Sharp-Bilinear-Shaders
• https://github.com/libretro/glsl-shaders
• https://www.codeproject.com/articles/236394/bi-cubic-and-bi-linear-interpolation-with-glsl
• https://www.shadertoy.com/view/XsjSzR

[ultraq]

One thing I'm finding out with the scanline shaders out there: they only work if all pixels are the same height (I mean, why not? 😅), but I have the situation here where I'm adjusting texture heights for the difference in aspect ratios, so some scanlines are coming through thicker than others. And since it's not a texture, it's not getting the same hardware filtering so those thick lines aren't lightened to try reduce their apparent thickness 🤔