llama : add 18-layer model type for Gemma 3-270m (#15319) #10

Job	Run time
macOS-arm64	2m 34s
windows-hip (radeon, gfx1100;gfx1101;gfx1102;gfx1030;gfx1031;gfx1032)	59m 50s
ios-xcode-build	19m 11s
windows-sycl	10m 35s
macOS-x64	7m 42s
windows-cpu (arm64)	2m 36s
ubuntu-22-cpu (x64, ubuntu-22.04)	4m 1s
windows (vulkan, x64, -DGGML_VULKAN=ON, ggml-vulkan)	3m 58s
ubuntu-22-vulkan	7m 0s
windows (opencl-adreno, arm64, -G "Ninja Multi-Config" -D CMAKE_TOOLCHAIN_FILE=cmake/arm64-window...	1m 9s
windows-cuda (12.4)	47m 5s
windows-cpu (x64)	3m 36s
release	1m 16s
	2h 50m 33s

Provide feedback