[P1] event-exporter: bounded retries with backoff for Cloud Logging writes

[erain]

Problem

writer.Write retries indefinitely with a fixed 10s delay. During API outages or throttling, this can stall throughput and cause backlog growth without a clear signal or drop policy.

Proposed optimization

Introduce bounded retry with exponential backoff + jitter, and expose metrics for retry count and dropped batches. Optionally apply per-status handling (e.g., retry 429/5xx, fail fast on 400). This makes failure behavior predictable and prevents infinite backlog under sustained failure.

Notes / references

• Cloud Logging WriteEntries API: https://cloud.google.com/logging/docs/reference/v2/rest/v2/entries/write
• Kubernetes client-go workqueue rate limiters (example for backoff): https://pkg.go.dev/k8s.io/client-go/util/workqueue

Acceptance criteria

• Retries use exponential backoff with jitter and a max retry time or count.
• Metrics expose retry attempts and dropped batches.
• Behavior for 400 vs 429/5xx is explicit and documented.