Implement linux NUMA detection

If the system has multiple numa nodes, enable numa support in llama.cpp
If we detect numactl in the path, use that, else use the basic "distribute" mode.

llm/server.go

@@ -256,8 +256,14 @@ func NewLlamaServer(gpus gpu.GpuInfoList, model string, ggml *GGML, adapters, pr
 		params = append(params, "--mlock")
 	}
 
-	if opts.UseNUMA {
-		params = append(params, "--numa")
+	if gpu.IsNUMA() {
+		numaMode := "distribute"
+		if runtime.GOOS == "linux" {
+			if _, err := exec.LookPath("numactl"); err == nil {
+				numaMode = "numactl"
+			}
+		}
+		params = append(params, "--numa", numaMode)
 	}
 
 	params = append(params, "--parallel", strconv.Itoa(numParallel))