switch to llama.cpp submodule for up to date llama

.gitmodules

@@ -0,0 +1,3 @@
+[submodule "llama.cpp"]
+	path = llama.cpp
+	url = https://github.com/ggml-org/llama.cpp