feat(batch): retryStrategy + timeout (batch 5)

[vieiralucas]

Summary

AWS Batch batch 5 — retry strategies + job timeouts.

• spawn_status_sync loops over attempts: on a non-zero container exit, if retryStrategy.attempts remain it records the failed attempt under attempts[] and re-launches a fresh ECS task (status RUNNABLE → STARTING), else fails the job. Attempts clamped to AWS's 1-10.
• timeout.attemptDurationSeconds caps each attempt's poll loop; an overrunning container fails the job with "Job attempt duration exceeded timeout".
• Extracted set_job_terminal for the SUCCEEDED/FAILED write.

Test plan

• Docker-gated e2e: retry_strategy_reattempts_a_failing_job (exit 4, attempts=2 → FAILED with one recorded retry + exitCode 4) and timeout_fails_an_overrunning_job (sleep 60, attemptDurationSeconds=2 → FAILED "timeout"). Both pass locally.
• Existing success/fail/array/dependsOn e2e unaffected; cargo clippy -p fakecloud-batch -p fakecloud-e2e --all-targets -- -D warnings clean. Docs updated.

Next (final Batch batch)

CloudFormation provisioner for AWS::Batch::ComputeEnvironment/JobQueue/JobDefinition (write-through + snapshot hook) + Terraform acceptance-test coverage.

---

Summary by cubic

Adds AWS Batch retry strategies and per-attempt timeouts to real job execution. Failed attempts re-launch up to retryStrategy.attempts (1–10) with attempts recorded; overruns fail with a clear timeout reason. Docs updated and e2e coverage added.

• New Features

  • Honor retryStrategy.attempts: on non‑zero exit, record in attempts[] and re-launch a fresh ECS task (RUNNABLE → STARTING) until attempts are exhausted.
  • Honor timeout.attemptDurationSeconds: cap each attempt; on overrun, fail with "Job attempt duration exceeded timeout".
  • Set RUNNING on first task start; terminal status writes keep real container.exitCode and statusReason.

• Refactors

  • Extracted set_job_terminal helper for SUCCEEDED/FAILED writes.

^{Written for commit b670f64. Summary will update on new commits.}