[Metal][DXC] Concurrent writing of `uint4` components test failure

[inbelic]

As illustrated in the ComponentDataRace test, Metal fails to correctly copy over the data when written concurrently across waves:

2025-09-24T18:54:33.7663560Z ******************** TEST 'OffloadTest-mtl :: WaveOps/ComponentDataRace.test' FAILED ********************
2025-09-24T18:54:33.7673620Z Exit Code: 1
2025-09-24T18:54:33.7673970Z 
2025-09-24T18:54:33.7674170Z Command Output (stdout):
2025-09-24T18:54:33.7674420Z --
2025-09-24T18:54:33.7674640Z Using Metal API
2025-09-24T18:54:33.7674870Z Created compute commands.
2025-09-24T18:54:33.7675090Z 
2025-09-24T18:54:33.7675220Z --
2025-09-24T18:54:33.7676900Z Command Output (stderr):
2025-09-24T18:54:33.7677860Z --
2025-09-24T18:54:33.7679250Z /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/bin/split-file /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/OffloadTest/test/WaveOps/ComponentDataRace.test /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp # RUN: at line 64
2025-09-24T18:54:33.7681620Z + /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/bin/split-file /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/OffloadTest/test/WaveOps/ComponentDataRace.test /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp
2025-09-24T18:54:33.7684220Z "/Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/DXC/build/bin/dxc-3.7" -metal -T cs_6_5 -Fo /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp.o /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp/source.hlsl # RUN: at line 65
2025-09-24T18:54:33.7686860Z + /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/DXC/build/bin/dxc-3.7 -metal -T cs_6_5 -Fo /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp.o /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp/source.hlsl
2025-09-24T18:54:33.7699530Z /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/bin/offloader -debug-layer /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp/pipeline.yaml /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp.o # RUN: at line 66
2025-09-24T18:54:33.7702360Z + /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/bin/offloader -debug-layer /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp/pipeline.yaml /Users/cbieneman/actions-runner/_work/offload-test-suite/offload-test-suite/llvm-project/build/tools/OffloadTest/test/mtl/WaveOps/Output/ComponentDataRace.test.tmp.o
2025-09-24T18:54:33.7704490Z 2025-09-24 13:54:33.574 offloader[39126:172017742] *** -[NSAutoreleasePool addObject:]: Do not use this instance method on specific pools -- just use -autorelease instead.
2025-09-24T18:54:33.7704960Z Test failed: ExpectedOut
2025-09-24T18:54:33.7705130Z Comparison Rule: BufferExact
2025-09-24T18:54:33.7705280Z Expected:
2025-09-24T18:54:33.7705380Z ---
2025-09-24T18:54:33.7705490Z Name:            ExpectedOut
2025-09-24T18:54:33.7705640Z Format:          UInt32
2025-09-24T18:54:33.7705760Z Channels:        4
2025-09-24T18:54:33.7705890Z Data:            [ 1, 2, 3, 4 ]
2025-09-24T18:54:33.7706040Z OutputProps:
2025-09-24T18:54:33.7706510Z   Height:          0
2025-09-24T18:54:33.7706640Z   Width:           0
2025-09-24T18:54:33.7706760Z   Depth:           0
2025-09-24T18:54:33.7706880Z ...
2025-09-24T18:54:33.7706970Z Got:
2025-09-24T18:54:33.7707060Z ---
2025-09-24T18:54:33.7707160Z Name:            Out
2025-09-24T18:54:33.7707290Z Format:          UInt32
2025-09-24T18:54:33.7707410Z Channels:        4
2025-09-24T18:54:33.7707540Z Data:            [ 1, 0, 0, 0 ]
2025-09-24T18:54:33.7707680Z OutputProps:
2025-09-24T18:54:33.7707790Z   Height:          0
2025-09-24T18:54:33.7708050Z   Width:           0
2025-09-24T18:54:33.7708170Z   Depth:           0
2025-09-24T18:54:33.7708290Z ...
2025-09-24T18:54:33.7708450Z Full Hex 64bit representation of Expected Buffer Values:
2025-09-24T18:54:33.7708670Z [ 0x1, 0x2, 0x3, 0x4 ]
2025-09-24T18:54:33.7708840Z Full Hex 64bit representation of Actual Buffer Values:
2025-09-24T18:54:33.7709890Z [ 0x1, 0x0, 0x0, 0x0 ]
2025-09-24T18:54:33.7710000Z 
2025-09-24T18:54:33.7710010Z 
2025-09-24T18:54:33.7710050Z --
2025-09-24T18:54:33.7710120Z 
2025-09-24T18:54:33.7710160Z ********************