[Feature] The headset's Forward pass and temporal resolve do less per pixel, and the sun casts in one cascade

[miogds]

Stacked on feature/xr_frame_cost: the headset's Forward pass and temporal resolve do less per pixel, in the shaders, with the kernel libraries rebuilt (buildkernels.sh, every platform); and the sun made cheap for the projects that switch it on. Four commits.

Why

After the Swift-side cuts the heavy configuration held 90 Hz with 8 ms of GPU a frame, 5.5 of it the Forward pass's fragment work. The viewer switches its sun off, half its materials lack a map, the room probe is never rotated, and most of a passthrough frame is cleared; the shaders paid for all four regardless. With the sun on, the frame no longer fitted at 100 Hz.

What changes

• The sun by its intensity. The cascade shadow taps and the sun's Cook-Torrance BRDF run only when the sun's intensity is not zero, in the forward shader, both light passes and the transparency shader, on a per-frame constant (a uniform branch). Any other intensity takes the old path in the old order of products, bit for bit.
• Each material map sampled only when the material has it, the normal map's decode and tangent transform included; the flags are per-draw constants, so the implicit-derivative samples stay in uniform control flow. The reviewers put this at nothing measurable; it stays as a cleanup.
• Once per pixel instead of twice: the view vector, the environment's rotation in radians, the BRDF lookup table's pair (both reflectances derive from it); the rotation of the lookup direction is skipped at angle zero, the room probe's case. The normalize is kept, because the simulator's light pass hands in a G-buffer normal as stored and the equirectangular latitude's asin is not scale-invariant; that asin is clamped now, which removes a pre-existing chance of a NaN pixel at the poles.
• The temporal resolve leaves early for a pixel with no history or at the clear depth, before the rate-map decodes, the reprojection and the velocity fetch. The foveation rates are read as texels per screen pixel over the same screen interval from both maps, a second decode of each, instead of derivatives across the quad, which forbade any early return: the ratio is exactly 1 wherever the two maps agree, zone boundaries included.
• The sun's cascades are a setting, setRendering(.shadowCascades(1 or 2)): one on the headset by default, where a room or a tabletop never shows the distance the second covers; with maxShadowCastingDistance set to the scene's size one cascade is sharp on it. The count flows through the splits, the fitting, both shadow passes and the shader's uniforms.
• A sun that casts no shadow shadows nothing. When a sun stops casting the cascades are no longer drawn, but the shader kept sampling the last maps; its matrices now land every point outside the maps, which the shader reads as lit.

Twelve reviewers (Metal semantics, picture equivalence, performance, three per cut) tried to refute each shader cut; what they found is folded in.

On the device

The sun off, 100 Hz: Forward 8.02 → 6.56 ms (the fragment part 5.51 → 4.15), TAA 2.06 → 1.92 (the empty frame 1.81 → 1.47), the frame 6.8 ms against the 10 ms refresh, no late frame after the placement. The sun on, one cascade: Shadows 0.85 → 0.28 ms, the sun about 0.9 ms in all; full rate at 90 Hz.

Tests

The material shading, per-fragment lighting, glass, probe, post-effects and temporal tests pass on the rebuilt library (148 in the targeted run); the cascade setting, its headset default and that one cascade still casts on a floor are covered by render tests, the non-casting matrices by a unit test.

[miogds]

Closed without merging here: the Forward pass and resolve doing less per pixel (feature/xr_shader_cost at eac4e3e) is on develop since #76 (merge 0b6348e of feature/xr_shader_cost at eac4e3e, which carries #73–#75), and develop now also holds the head of untoldengine#1359 through #78 and #79. The branch stays for the upstream pull request once untoldengine#1359 lands.