feat(otel): Add checkpoint restart-recovery integration test

[louisall]

Summary

End-to-end integration test verifying that file_storage checkpoint persistence prevents log loss across agent pod restarts.

Test Flow

1. Confirms agent is tailing nginx-test (warmup marker arrives in CW Logs)
2. Emits 20 numbered markers via /proc/1/fd/1 to the container log stream
3. Kills the agent pod mid-sequence (after marker 10)
4. Emits remaining markers while agent is restarting
5. Waits for agent restart + log propagation
6. Queries CW Logs: asserts all 20 markers arrive exactly once

Validates: no gaps (checkpoint resumes correctly) and no duplicates (offset is accurate).

Dependencies

• Base: PR feat(otel): Add OTLP logs integration test framework #697 (feat/otel-logs-integration-tests) — requires the OTLP logs test framework (logsClient, appLogGroup(), etc.)
• Helm chart: helm-charts PR #313 — the file_storage config this test validates

Design Decisions

• Uses existing nginx-test deployment (already deployed by terraform for metrics tests) — no additional terraform resources needed
• Writes to /proc/1/fd/1 to inject markers into the container log stream (not exec stdout)
• Targets standard 10.0.x.x nodes (skips Fargate/special nodes)
• Runs after metrics tests when the agent has been tailing nginx-test for several minutes

Results (manual run on metricsv2-testing cluster)

marker-001: 1
...
marker-020: 1
PASS: all 20 markers delivered exactly once