Fix: Kura gateway no longer stuck FailedScheduling on bare-metal cache nodes

The Kura gateway Helm release now tolerates the tuist.dev/kura-cache=true:NoSchedule taint on bare-metal cache nodes, so the host-network regional ingress-nginx gateway can schedule onto the same dedicated cache box as the KuraInstance cache pods. Without this, on bare-metal regions (Dedibox/OVH-style) the regional gateway sat in FailedScheduling because it had to bind the box’s public IP via nodeSelector/hostNetwork but carried no toleration, leaving the region without ingress and the public cache endpoint unreachable. The gateway spec and pod template now pass the same tolerations(region) the cache pods already use, so cloud/LB regions remain unchanged and bare-metal regions come up correctly.