Rename to gpu-turnstile; swap port roles

The proxy now listens on the standard service ports (Ollama :11434,
ComfyUI :8188) and the actual services move one port up (:11435, :8189).
Metric prefix is now gpu_turnstile_.

internal/proxy/proxy_test.go

@@ -40,13 +40,13 @@ func (r *recorder) index(e string) int {
 	return -1
 }
 
-// fakes wires up fake Ollama and ComfyUI upstreams plus the gpulock server.
+// fakes wires up fake Ollama and ComfyUI upstreams plus the gpu-turnstile server.
 type fakes struct {
 	rec       *recorder
 	ollama    *httptest.Server
 	comfy     *httptest.Server
-	server    *httptest.Server // Ollama-facing gpulock listener
-	comfySrv  *httptest.Server // ComfyUI-facing gpulock listener
+	server    *httptest.Server // Ollama-facing gpu-turnstile listener
+	comfySrv  *httptest.Server // ComfyUI-facing gpu-turnstile listener
 	freeCh    chan struct{}
 	chatCh    chan struct{}
 	historyMu sync.Mutex
@@ -279,7 +279,7 @@ func TestStreamingNotBuffered(t *testing.T) {
 	}
 	defer resp.Body.Close()
 
-	// If gpulock buffered the stream, this read would never complete while
+	// If gpu-turnstile buffered the stream, this read would never complete while
 	// the gate is closed.
 	line, err := bufio.NewReader(resp.Body).ReadString('\n')
 	if err != nil {
@@ -321,11 +321,11 @@ func TestHealthzAndMetrics(t *testing.T) {
 	resp.Body.Close()
 	text := string(body)
 	for _, want := range []string{
-		`gpulock_state{state="idle"} 1`,
-		"gpulock_llm_inflight 0",
-		"gpulock_image_jobs_total 0",
-		`gpulock_lock_wait_seconds_bucket{kind="llm",le="+Inf"} 0`,
-		"gpulock_unload_seconds_count{} 0",
+		`gpu_turnstile_state{state="idle"} 1`,
+		"gpu_turnstile_llm_inflight 0",
+		"gpu_turnstile_image_jobs_total 0",
+		`gpu_turnstile_lock_wait_seconds_bucket{kind="llm",le="+Inf"} 0`,
+		"gpu_turnstile_unload_seconds_count{} 0",
 	} {
 		if !strings.Contains(text, want) {
 			t.Fatalf("metrics missing %q", want)