fix: update deployment strategy to Recreate for tts-gateway to prevent GPU deadlock

manifests/home-services/tts-gateway.yaml

@@ -7,6 +7,12 @@ metadata:
     app: tts-gateway
 spec:
   replicas: 1
+  # nik-gpu only has 1 nvidia.com/gpu allocatable - the default RollingUpdate
+  # strategy tries to schedule the new pod before killing the old one, which
+  # deadlocks forever waiting for a GPU that's still held by the pod it's
+  # replacing. Recreate kills the old pod first instead.
+  strategy:
+    type: Recreate
   selector:
     matchLabels:
       app: tts-gateway