Advance native Windows RR integration, kernel tooling and local packaging

Implement native NGX feature-13 RR through a LUID-matched D3D12/Vulkan
shared-memory and timeline-fence bridge. Add owned parameter snapshots,
GPU resource lifetime handling, inherited/enhanced barriers, optional
guides, sampled input formats and depth-plane support with stencil
preservation. Separate RR initialization from the SR CUDA runtime and
expose explicit backend/model selection through player configuration.

Add synthetic bridge, snapshot and NGX lifecycle harnesses covering
camera/history/reset, parameter mutation, delayed submission, early
release, capability gating and unsupported-input rejection.

Add exact gfx11 D/E model generation, verified caches, versioned local
deployment with configuration backups, and isolated portable model
tools with pinned Python, NumPy and glslang dependencies.

Extend Windows K benchmarking and private native-cache selection.
Fix LLVM 22 emitter/helper compatibility and preserve exact-output
validation for local output-kernel optimization.

Add native FSR RR provider diagnostics, local development package
refresh tooling, installer/checker updates, FidelityFX headers and
Windows parity documentation.

Keep NVIDIA binaries, derived models and private optimized kernels out
of distributable packages. Native RR evidence remains synthetic;
game validation, FSR RR dispatch, installer integration and Linux
performance parity remain incomplete.

scripts/windows/rr-model-cache.py

@@ -0,0 +1,72 @@
+"""Stamp or verify local RR build artifacts against sources, tools and DLL identity."""
+import argparse
+import hashlib
+import json
+from pathlib import Path
+import sys
+
+
+def sha(path):
+    return hashlib.sha256(path.read_bytes()).hexdigest()
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument('action', choices=('write', 'check'))
+    parser.add_argument('--model', type=Path, required=True)
+    parser.add_argument('--dll', type=Path, required=True)
+    parser.add_argument('--preset', choices=('D', 'E'), required=True)
+    parser.add_argument('--glslang', type=Path, required=True)
+    parser.add_argument('--bash', type=Path, required=True)
+    args = parser.parse_args()
+    repo = Path(__file__).resolve().parents[2]
+    model = args.model.resolve()
+    if not model.is_relative_to((repo / 'build').resolve()):
+        parser.error('model cache must stay under ignored build/')
+    import numpy
+    sources = {}
+    for directory in ('engine', 'vulkan', 'kernels/tools', 'scripts/windows'):
+        for path in sorted((repo / directory).rglob('*')):
+            if path.is_file() and path.suffix in ('.py', '.ps1', '.sh', '.comp', '.glsl', '.h', '.inc', '.txt'):
+                sources[path.relative_to(repo).as_posix()] = sha(path)
+    identity = dict(schema=1, preset=args.preset, architecture='gfx1101', arithmetic='exact-gfx11',
+                    numpyVersion=numpy.__version__, sources=sources,
+                    tools={name: sha(path) for name, path in (
+                        ('dll', args.dll), ('glslang', args.glslang),
+                        ('bash', args.bash), ('python', Path(sys.executable)))})
+    stamp = model / 'windows-rr-cache.json'
+    products = {path.name: sha(path) for path in sorted(model.iterdir())
+                if path.is_file() and path.name not in (stamp.name, stamp.name + '.tmp')}
+    for name in ('rr.txt', 'model.json', 'enc0_m.spv', 'dec0_m.spv', 'hkpn.spv', 'enc0.args', 'hkpn.args'):
+        if name not in products:
+            raise RuntimeError(f'incomplete model: {name}')
+    manifest = (model / 'rr.txt').read_text().splitlines()
+    if (not manifest or manifest[0] != 'D4RR0001' or
+            f'preset {4 if args.preset == "D" else 5}' not in manifest or
+            'arithmetic gfx11' not in manifest):
+        raise RuntimeError('RR manifest preset/arithmetic does not match requested cache identity')
+    stages = [line.split() for line in manifest if line.startswith('stage ')]
+    expected_stages = {'enc0', 'enc1', 'enc2', 'enc3', 'enc4', 'enc5',
+                       'dec4', 'dec3', 'dec2', 'dec1', 'dec0', 'hkpn'}
+    if len(stages) != 12 or {fields[1] for fields in stages} != expected_stages:
+        raise RuntimeError('RR stage list is incomplete or duplicated')
+    for fields in stages:
+        for name in (fields[2], fields[4]):
+            if name != '-' and name not in products:
+                raise RuntimeError(f'RR stage artifact missing or outside model directory: {name}')
+    if args.action == 'check':
+        cached = json.loads(stamp.read_text())
+        if cached.get('identity') != identity:
+            raise RuntimeError('RR cache identity changed; use a new output directory to rebuild')
+        if cached.get('products') != products:
+            raise RuntimeError('RR cache artifacts changed; use a new output directory to rebuild')
+        print(f'RR_CACHE_HIT preset={args.preset} artifacts={len(products)} model={model}')
+    else:
+        temporary = stamp.with_name(stamp.name + '.tmp')
+        temporary.write_text(json.dumps(dict(identity=identity, products=products), indent=2) + '\n')
+        temporary.replace(stamp)
+        print(f'RR_CACHE_WRITTEN preset={args.preset} artifacts={len(products)} model={model}')
+
+
+if __name__ == '__main__':
+    main()