fix(applications): wire newArgocdClientFn in Application controller connector

[gravufo]

Summary

Fixes #372.

Setup() for the Application controller built its connector without setting newArgocdClientFn, leaving it nil. Every Connect() call then invokes that nil function pointer inside NewExternal(), panicking on every reconcile. controller-runtime recovers the panic and requeues with backoff instead of crashing, so the resource silently never gets its status populated — no error, no event, nothing beyond a bare Reconciling log line forever.

This brings applications/controller.go in line with projects/controller.go and cluster/controller.go, which already wire their respective newArgocdClientFn (projects.NewProjectServiceClient, etc.) into the connector.

Change

func Setup(mgr ctrl.Manager, o xpcontroller.Options) error {
	return SetupWithExternalConnector(mgr, o, &connector{
		kube:              mgr.GetClient(),
		newArgocdClientFn: applications.NewApplicationServiceClient,
	})
}

Test plan

• go build ./pkg/controller/cluster/applications/...
• go vet ./pkg/controller/cluster/applications/...
• Verified in a live cluster: before the fix, Application managed resources never populated .status and the controller logged only a repeating bare Reconciling line with no follow-up (unlike Cluster/Project, which log the full observe/update sequence). Manually patching .status directly on the object and observing it survive multiple poll cycles untouched confirmed the controller was never reaching any status-updating code path.