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.go

@@ -1,4 +1,4 @@
-// Package proxy contains the HTTP handlers for both gpulock listeners:
+// Package proxy contains the HTTP handlers for both gpu-turnstile listeners:
 // reverse proxies to Ollama and ComfyUI with GPU lock arbitration in front
 // of the endpoints that load models.
 package proxy
@@ -43,7 +43,7 @@ type Config struct {
 	WarmModel      string
 }
 
-// Server serves both gpulock listeners.
+// Server serves both gpu-turnstile listeners.
 type Server struct {
 	cfg Config
 	log *slog.Logger