fix(gateway) handle NodePort addresses

[rainest]

What this PR does / why we need it:

Ensure that the controller's desired addresses for Services with no external addresses (such as NodePorts) match the Gateway representation of no addresses. This ensures that the address update check does not attempt to proceed with an unnecessary update every reconcile, which would prevent the Gateway from becoming ready.

Add an E2E test that confirms Gateways work with NodePorts.

Which issue this PR fixes:

Fix #2808
Fix #2823

Special notes for your reviewer:

E2E run: https://github.com/Kong/kubernetes-ingress-controller/actions/runs/2871403290

PR Readiness Checklist:

Complete these before marking the PR as ready to review:

• the CHANGELOG.md release notes have been updated to reflect any significant (and particularly user-facing) changes introduced by this PR this was a bug introduced by unreleased code from Update Gateway status on ReferenceGrant changes #2797

[rainest]

This is currently affected by the failure that #2824 will fix. Draft waiting for that; this needs to be rebased after that's merged.

2022-08-16T20:24:15.5342094Z     helpers_test.go:290: 
2022-08-16T20:24:15.5343441Z         	Error Trace:	/home/runner/work/kubernetes-ingress-controller/kubernetes-ingress-controller/test/e2e/helpers_test.go:290
2022-08-16T20:24:15.5345023Z         	            				/home/runner/work/kubernetes-ingress-controller/kubernetes-ingress-controller/test/e2e/features_test.go:337
2022-08-16T20:24:15.5345626Z         	Error:      	Condition never satisfied
2022-08-16T20:24:15.5346129Z         	Test:       	TestDeployAllInOneDBLESSGateway
2022-08-16T20:24:15.5347559Z     features_test.go:286: cleaning up environment for cluster 14e300ea-8260-41ba-9b37-7493e874583f
2022-08-16T20:24:17.0472065Z --- FAIL: TestDeployAllInOneDBLESSGateway (396.49s)

[rainest]

On second thought, looking at that failure, configuring the route to strip paths makes more sense than intentionally causing a 404 and scraping the server header, so I've added that change here.

[randmonkey]

On second thought, looking at that failure, configuring the route to strip paths makes more sense than intentionally causing a 404 and scraping the server header, so I've added that change here.

OK, I did not find that we have an annotation to do the strip path of HTTPRoute. I think we do not need to put #2824 as a blocker of this.