Capture abnormal runner death logs before Pod reap

Operators running the Tuist runners controller can now diagnose mid-job Linux runner deaths that previously appeared as “self-hosted runner lost communication”. When a runner Pod ends abnormally, the controller captures the runner container’s final log tail and re-emits it to its own durable stdout before the Pod is reaped, so the trail survives the cleanup. The Helm chart 0.8.0 includes the updated Role that grants the controller pods/log access. Operators can query Loki for runner death log captured to inspect the captured trail. There is no runtime behavior change for normal jobs.