Capture runner death logs before Pods are reaped

When a self-hosted runner Pod dies abnormally mid-job, the runners controller now fetches the runner container’s recent log tail and re-prints it to its own stdout before the Pod is deleted. This preserves a diagnostic trail that used to disappear with the Pod, making infrastructure-level runner deaths easier to investigate. Operators can query these captures in Loki with {pod=~".*runners-controller.*"} |= "runner death log captured".