[Rendering] fix: guard the bright pass against infinities and NaNs

[Nicogo1705]

PR Details

Split out of #3387, per review.

The bright filter's output is downsampled and blurred through a pyramid of half-float targets. A value above 65504 becomes an infinity there, the first zero blur weight turns it into a NaN, and the NaN spreads across the kernel as a black block the size of the mip it happened on. Luma already floored negatives; nothing capped the other end.

• The bright filter and the range compressor clamp to 32000, half a half-float's ceiling, through a shared HalfPrecisionUtils.ClampToSafeRange.
• Written as color > 0.0 ? min(color, MaxSafeValue) : 0.0: a per-component select, so no branch, and a NaN fails the comparison and becomes 0 on every backend. min/max on a NaN is undefined in SPIR-V, so the earlier min(max(...)) form only worked on Direct3D.

Verified on Direct3D 11 with the voxel GI demo, bloom on: no black blocks, same image otherwise.

Related Issue

#3387

Types of changes

• Bug fix (non-breaking change which fixes an issue)

Checklist

• All new and existing tests passed.
• I have built and run the editor to try this change out.

---

_{⚠️ Written with AI assistance (Claude Code), driven and verified by me; needs a real review. 🤖 Generated with Claude Code}