rr: add FSR Ray Regeneration, exact RDNA3 inference and motion experiments

Add an experimental FSR Ray Regeneration backend for D3D12 RR requests:
- Convert depth, motion, normals and albedos into AMD denoiser inputs.
- Split and recompose diffuse/specular signals, then run Super Sampling.
- Add stability/detail controls, demodulation guards and halo limiting.
- Handle AMD provider discovery under NVIDIA GPU spoofing.
- Add runtime probes, synthetic reference scenes and per-dispatch profiling.

Add exact gfx11 DLSS-RR inference using f16 cooperative matrices:
- Reproduce FP8 conversions, accumulation and camera reprojection arithmetic.
- Widen model weights and avoid requiring VK_EXT_shader_float8.
- Extend model building, replay tools and descriptor metadata.

Expand motion preprocessing and experimental AW2 native-motion support:
- Add tiled LDS dilation, padded allocation handling and thin-only sky fill.
- Order dilation with input waits and NGX on the selected stream.
- Add jitter-aligned native composition, hybrid routing and wire history repair.
- Add prepared masks, sparse repair and independent motion/repair controls.

Add optional fused R11G11B10 input/output conversion, sampled stage timings
and profile summarization. Extend configuration, packaging, documentation,
FidelityFX headers and CPU/GPU regression coverage.

tests/test_vram_profile_timing.py

@@ -0,0 +1,42 @@
+"""Conversion labels must belong to the older Vulkan query slot, not this frame."""
+from pathlib import Path
+import subprocess
+import tempfile
+import unittest
+
+ROOT=Path(__file__).resolve().parents[1]
+
+
+class VramProfileTimingTests(unittest.TestCase):
+    def test_older_slot_mode_labels(self):
+        source=(ROOT/'tools/d4r_nvngx_shim.cpp').read_text()
+        start=source.index('static void read_profile_timestamps(Feature& feature, uint32_t frame, FrameTiming& timing)\n{')
+        end=source.index('\n}\n',start)+3
+        program=r'''
+#include <cassert>
+#include <cstdint>
+constexpr unsigned kOutputSlots=4;
+struct FrameTiming {double vkInput=-1,vkDlssWait=-1,vkOutput=-1,vkTotal=-1;int vkOutputFused=-1,vkInputFused=-1;};
+struct Feature {uint64_t values[16]={};volatile uint64_t* timestampValues=values;double timestampTicksPerMs=10;
+ uint8_t timestampOutputFused[4]={},timestampInputFused[4]={};bool recordOutputFused=true,recordInputFused=true;};
+''' + source[start:end] + r'''
+int main() {
+ Feature f;FrameTiming old;
+ f.values[0]=10;f.values[1]=20;f.values[2]=50;f.values[3]=60;
+ f.timestampOutputFused[0]=0;f.timestampInputFused[0]=1;
+ read_profile_timestamps(f,300,old); // frame 296 used transfer output, despite current mode switching on
+ assert(old.vkOutputFused==0 && old.vkInputFused==1 && old.vkTotal==5);
+ assert(!f.recordOutputFused && !f.recordInputFused && f.values[0]==0);
+ f.values[0]=70;f.values[1]=80;f.values[2]=110;f.values[3]=120;
+ f.timestampOutputFused[0]=1;f.timestampInputFused[0]=0;FrameTiming next;
+ read_profile_timestamps(f,304,next);
+ assert(next.vkOutputFused==1 && next.vkInputFused==0);
+ FrameTiming unavailable;read_profile_timestamps(f,308,unavailable);
+ assert(unavailable.vkOutputFused==-1 && unavailable.vkInputFused==-1 && unavailable.vkTotal==-1);
+}
+'''
+        with tempfile.TemporaryDirectory() as temp:
+            runner=Path(temp)/'timing'
+            subprocess.run(['c++','-std=c++17','-Wall','-Wextra','-Werror','-x','c++','-',
+                            '-o',str(runner)],input=program,text=True,check=True,capture_output=True)
+            subprocess.run([str(runner)],check=True,timeout=5)