[doc] Add GKE Gateway example for ingress

[lalitc375]

Description

This PR updates the KubeRay ingress documentation to include a comprehensive guide on using the GKE Gateway API to access a Ray cluster.

Specifically, this PR accomplishes the following:

• Adds a new 5th example section detailing GKE Gateway API support.
• Provides a complete YAML configuration (ray-cluster-gke-gateway.yaml) for setting up a Gateway (using the gke-l7-rilb internal gateway class) and an HTTPRoute to route traffic to the raycluster-kuberay-head-svc on port 8265.
• Includes step-by-step instructions (Steps 1 through 8) covering everything from installing the KubeRay operator via Helm, applying the gateway configurations, to fetching the allocated internal IP for accessing the Ray Dashboard.

Related issues

NONE

Additional information

NONE

[gemini-code-assist]

Code Review

This pull request adds documentation for GKE Gateway API support, including a new guide with prerequisites and step-by-step instructions for deploying Gateway and HTTPRoute resources. Feedback suggests improving the guide by using kubectl wait for status checks and clarifying the accessibility limitations of internal IP addresses, including a kubectl port-forward alternative.

[gemini-code-assist]

Since the example uses the gke-l7-rilb (Regional Internal Load Balancer) class, the allocated IP address is internal to your VPC. It will not be accessible directly from a browser on your local machine unless you have a VPN, a proxy, or are using a jump box within the same VPC. It would be helpful to clarify this limitation to avoid user confusion when the dashboard is not reachable.

Suggested change

	# Step 7: Check Ray Dashboard by visiting the allocated internal IP in your browser. (In this example, it is 10.0.1.15)
	# Step 7: Check Ray Dashboard. Since this is an internal Gateway, the IP is only accessible from within the VPC. To access it from your local machine, you can use a VPN, a proxy, or for temporary testing, use `kubectl port-forward`:
	# kubectl port-forward service/raycluster-kuberay-head-svc 8265:8265

[gemini-code-assist]

Instead of waiting "a few minutes" manually, you can use kubectl wait to programmatically check when the Gateway has been assigned an IP and is ready.

Suggested change

	# Step 6: After a few minutes, GKE allocates an internal IP for the gateway. Check it using:
	kubectl get gateway ray-cluster-gateway
	# Step 6: Wait for GKE to allocate an internal IP and program the Gateway.
	kubectl wait --for=condition=Programmed gateway/ray-cluster-gateway --timeout=5m
	kubectl get gateway ray-cluster-gateway

[edoakes]

@andrewsykim PTAL

[spencer-p]

+1 to the gemini comments re kubectl wait, and that the internal LB address may not be reachable from the users' browser.

This guide is focused on exposing the Ray API -- if we want to leave a note for serving, this guide would be good to link to.

There is an aside about the risk of exposing the gateway to the internet -- I think there is maybe an opportunity to call out that the gateway can be configured to do authentication and authorization. Maybe just a note linking to Secure a Gateway? What do you think?

[lalitc375]

reply to @spencer-p
Addressed your comments.

[andrewsykim]

nit:

The following examples show how to use Ingress or Gateway to access your Ray clusters:

[lalitc375]

done