CUDA: add FP32 FlashAttention vector kernel #12040

Job	Run time
Push Docker image to Docker Hub (light, .devops/main.Dockerfile, linux/amd64,linux/arm64)	26m 22s
Push Docker image to Docker Hub (full, .devops/full.Dockerfile, linux/amd64,linux/arm64)	29m 18s
Push Docker image to Docker Hub (server, .devops/server.Dockerfile, linux/amd64,linux/arm64)	26m 17s
Push Docker image to Docker Hub (light-cuda, .devops/main-cuda.Dockerfile, linux/amd64)	25m 19s
Push Docker image to Docker Hub (full-cuda, .devops/full-cuda.Dockerfile, linux/amd64)	26m 36s
Push Docker image to Docker Hub (server-cuda, .devops/server-cuda.Dockerfile, linux/amd64)	24m 37s
Push Docker image to Docker Hub (light-rocm, .devops/main-rocm.Dockerfile, linux/amd64,linux/arm64)	10m 49s
Push Docker image to Docker Hub (full-rocm, .devops/full-rocm.Dockerfile, linux/amd64,linux/arm64)	10m 53s
Push Docker image to Docker Hub (server-rocm, .devops/server-rocm.Dockerfile, linux/amd64,linux/a...	10m 19s
Push Docker image to Docker Hub (light-intel, .devops/main-intel.Dockerfile, linux/amd64)	7m 26s
Push Docker image to Docker Hub (server-intel, .devops/server-intel.Dockerfile, linux/amd64)	7m 26s
	3h 25m 22s

Provide feedback