Use depth bias instead of physical offset for Text entity text

[ada-tv]

Improves the look of Text entities, since the physical offset was a global constant, it could make the text stick out pretty far on small Text entities and the slight parallax even on larger ones felt kinda weird to look at

I've tested this with some of my toy scripts and looking at the Text entities around the hub. Everything looks to be working fine with this change.

Closes #2006

Before

After

[JulianGro]

This PR causes z-fighting on my machine, both on GLES and Vulkan:

master:

This PR with Vulkan:

This PR with GLES:

This is all on the forward renderer though it also happens on deferred.
On a screenshot it isn't as obvious, but on master the text is just pixelated, while on this PR it flickers as if each letter was z-fighting.

[JulianGro]

On my workstation, this behaves a lot better. (Mesa 25.3.2-2 on Intel Arc A770)
OpenGL looks fine:

OpenGL_2026-01-14.mp4

Vulkan has very consistent z-fighting:

Vulkan_2026-01-14.mp4

OpenGL on master for comparison:

master_OpenGL_2026-01-14.mp4

On master, it is visible that things also move a little there, so that part is the same (this is on the forward renderer). It would be great to figure out what is happening on the Vulkan side.

[ada-tv]

I couldn't reproduce the z-fighting on GLES, but I was able to reproduce it on Vulkan:

This might just be that our Vulkan backend doesn't support depth biasing yet.