Emit an event when floating IP recreate fails midway

Context

The recreate path in ensureLoadBalancer (internal/provider/load_balancer.go) runs detach → delete → create when the allocator config changes. If create fails after delete succeeds, the service is left with no floating IP until the next reconcile — and nothing tells the operator what happened or why.

Scope

Emit a warning event (and structured log) identifying the window: the old IP was deleted due to a config change and recreation failed. Should be visible via kubectl describe svc.

Done when

A create failure after delete is diagnosable from service events without reading CCM source code.

Start in internal/provider/load_balancer.go at ensureLoadBalancer and trace the detach → delete → create path when allocator configuration changes. Inspect the branch where create fails after deletion, then run the relevant project checks and verify that the warning event and structured log identify the deleted old IP and failed recreation, with the event visible through kubectl describe svc.