Fix unschedulable Kura instances on bare-metal regions by using the shared regional gateway

Kura instances deployed to bare-metal (host-network) regions now route through the region’s shared gateway instead of a per-account dedicated gateway. Previously the server still tried to provision a dedicated ingress controller on the same single host, which couldn’t bind port 443 and left the instance unschedulable. Accounts on those regions now serve traffic correctly, while dedicated gateways on LoadBalancer regions are unchanged. To pick up the fix on an already-provisioned instance, perform a Destroy then Deploy in the dashboard so the manifest re-renders onto the shared ingress class.