fix: add retry logic to vpc-egress-gateway e2e curl connectivity check

The curl command in checkEgressAccess() used RunHostCmdOrDie which fails
immediately without retry. This causes flaky failures because VEG
Status.Ready only indicates control plane readiness, while the OVN
dataplane (NB→SB propagation, ovn-controller flow installation, BFD
session establishment) may still be converging.

Replace RunHostCmdOrDie with WaitUntil + RunHostCmd to retry for up to
30 seconds, matching the pattern already used by CheckPodEgressRoutes.

Signed-off-by: Mengxin Liu <liumengxinfly@gmail.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: oilbeater

test/e2e/vpc-egress-gateway/e2e_test.go

@@ -455,9 +455,15 @@ func checkEgressAccess(f *framework.Framework, namespaceName, svrPodName, image,
 		ginkgo.By("Checking connection from " + pod.Name + " to " + svrIP + " via " + protocol)
 		cmd := fmt.Sprintf("curl -q -s --connect-timeout 2 --max-time 2 %s/clientip", net.JoinHostPort(svrIP, svrPort))
 		ginkgo.By(fmt.Sprintf(`Executing %q in pod %s/%s`, cmd, pod.Namespace, pod.Name))
-		output := e2epodoutput.RunHostCmdOrDie(pod.Namespace, pod.Name, cmd)
-		clientIP, _, err := net.SplitHostPort(strings.TrimSpace(output))
-		framework.ExpectNoError(err)
+		var clientIP string
+		framework.WaitUntil(2*time.Second, 30*time.Second, func(_ context.Context) (bool, error) {
+			output, err := e2epodoutput.RunHostCmd(pod.Namespace, pod.Name, cmd)
+			if err != nil {
+				return false, nil
+			}
+			clientIP, _, err = net.SplitHostPort(strings.TrimSpace(output))
+			return err == nil, nil
+		}, fmt.Sprintf("curl %s from pod %s/%s", net.JoinHostPort(svrIP, svrPort), pod.Namespace, pod.Name))
 		framework.ExpectContainElement(expectedClientIPs, clientIP)
 	}
 }