[LLVMGPU] Use flat workgroup sizes in vector distribution #14034

Job	Run time
setup / setup	10s
Build Packages / Linux Release (x86_64)	15m 12s
Unit Test / Linux (x86_64)	2m 40s
Test AMD MI300 / test_mi300	1m 59s
Regression Test / test_models :: cpu_llvm_task	4m 42s
Regression Test / test_regression_suite :: cpu_llvm_task	10m 21s
Regression Test / test_models :: amdgpu_rocm_mi250_gfx90a	6m 40s
Test AMD MI250 / test_mi250	5m 10s
Test AMD W7900 / test_w7900	1m 48s
Regression Test / test_models :: amdgpu_rocm_mi300_gfx942	3m 21s
Regression Test / test_regression_suite :: amdgpu_rocm_mi250_gfx90a	10m 2s
Regression Test / test_models :: amdgpu_vulkan	1m 59s
Regression Test / test_regression_suite :: amdgpu_rocm_mi300_gfx942	8m 33s
Test Android / android_arm64	6m 1s
Test ONNX / test_onnx_models :: cpu_llvm_task	3m 4s
Test ONNX / test_onnx_ops :: cpu_llvm_sync	4m 14s
Test RISC-V 64 / riscv64	5m 6s
Test ONNX / test_onnx_ops :: amdgpu_rocm_rdna3	8m 22s
Test TensorFlow / Linux (x86_64)	2m 9s
Test ONNX / test_onnx_ops :: amdgpu_vulkan	1m 44s
	1h 43m 17s

Provide feedback