Revert "Optimize `fp8` `linalg_ext.attention` by rework Q@K scaling" #11289

Job	Run time
setup / setup	4s
Build Packages / Linux Release (x86_64)	21m 54s
Test AMD MI300	0s
Test AMD MI250	0s
Test AMD W7900	0s
Test NVIDIA T4	0s
Regression Test / test_models :: cpu_llvm_task	5m 10s
Regression Test / test_onnx :: cpu_llvm_sync	3m 7s
Regression Test / test_models :: amdgpu_rocm_mi250_gfx90a	4m 25s
Regression Test / test_onnx :: amdgpu_rocm_rdna3	7m 4s
Regression Test / test_models :: amdgpu_rocm_mi300_gfx942	4m 16s
Regression Test / test_onnx :: amdgpu_vulkan	1m 27s
Regression Test / test_models :: amdgpu_vulkan	2m 11s
Regression Test / test_regression_suite :: cpu_llvm_task	10m 33s
Regression Test / test_onnx :: nvidiagpu_cuda	5m 46s
Regression Test / test_regression_suite :: amdgpu_rocm_mi250_gfx90a	4m 27s
Regression Test / test_onnx :: nvidiagpu_vulkan	4m 57s
Regression Test / test_regression_suite :: amdgpu_rocm_mi300_gfx942	7m 52s
Unit Test / Linux (x86_64)	25s
test_android / cross_compile	5m 13s
Test TensorFlow CPU / Linux (x86_64)	2m 12s
	1h 31m 3s

Provide feedback