feat: Downsampling during camera movement

[aranega]

Summary

This PR adds XY downsampling for transparent layers (e.g. volumes) during camera movement (when the camera is actively rotating/panning). Transparent layers render into a smaller framebuffer (e.g. 4x fewer pixels), which is then composited back to the main canvas. When the camera stops, rendering returns to full resolution.

downsampling-2026-03-26_20.07.22.mp4

Implementation

DownsampledFramebuffer manages a frame buffer with a single RGBA color attachment (lazy resize with texture recreation), while DownsamplingCompositePass renders a fullscreen triangle that samples the downsampled texture. In the renderer beginDownsampling and endDownsampling wrap the renderTransparentLayers call. The framebuffer is owned by the renderer, created lazily on first use, and resized as needed. downsamplingFactor is a per-viewport config option (defaults to 1 = no downsampling). The cameraControls.isMoving signal controls when downsampling activates.

Tests & Checks

Manually tested with the volume_rendering and multi_viewport examples (both configured with downsamplingFactor: 4)

[seankmartin]

Overall looks good thanks! Wanted to note one difference in behaviour for the Idetik team to weigh in on. The others are very small comments