llamafile : improve moe prompt eval speed on cpu #11305

Job	Run time
Push Docker image to Docker Hub (light, .devops/main.Dockerfile, linux/amd64,linux/arm64)	27m 3s
Push Docker image to Docker Hub (full, .devops/full.Dockerfile, linux/amd64,linux/arm64)	27m 39s
Push Docker image to Docker Hub (server, .devops/server.Dockerfile, linux/amd64,linux/arm64)	24m 39s
Push Docker image to Docker Hub (light-cuda, .devops/main-cuda.Dockerfile, linux/amd64)	15m 26s
Push Docker image to Docker Hub (full-cuda, .devops/full-cuda.Dockerfile, linux/amd64)	14m 54s
Push Docker image to Docker Hub (server-cuda, .devops/server-cuda.Dockerfile, linux/amd64)	13m 26s
Push Docker image to Docker Hub (light-rocm, .devops/main-rocm.Dockerfile, linux/amd64,linux/arm64)	13m 41s
Push Docker image to Docker Hub (full-rocm, .devops/full-rocm.Dockerfile, linux/amd64,linux/arm64)	11m 16s
Push Docker image to Docker Hub (server-rocm, .devops/server-rocm.Dockerfile, linux/amd64,linux/a...	10m 37s
Push Docker image to Docker Hub (light-intel, .devops/main-intel.Dockerfile, linux/amd64)	6m 43s
Push Docker image to Docker Hub (server-intel, .devops/server-intel.Dockerfile, linux/amd64)	6m 59s
	2h 52m 23s

Provide feedback