fix(batch): ListJobs semantics, tag read/write parity, queue validation

[vieiralucas]

AWS Batch correctness cluster from the 2026-06-27 bug-hunt (Tier 1: findings 1.1-1.6).

ListJobs (crates/fakecloud-batch/src/service.rs)

• 1.1 Default to RUNNING when no jobStatus is given. AWS: "If you don't specify a status, only RUNNING jobs are returned." Previously returned every status -> pollers over-counted running jobs.
• 1.2 Honor the arrayJobId / multiNodeJobId selectors and require exactly one of jobQueue/arrayJobId/multiNodeJobId (AWS rejects otherwise). Enumerating an array job's children works now.
• 1.4 Paginate via maxResults (<=100) + opaque nextToken.
• 1.5 Emit the full JobSummary (jobArn/statusReason/startedAt/stoppedAt/container/arrayProperties/jobDefinition), not just jobId/name/status/createdAt.

Tags single-store (1.3)

Create wrote tags inline on the resource; TagResource/UntagResource wrote a separate per-ARN store; Describe* read inline only and ListTagsForResource read the store only — so the two diverged and terraform tag updates never converged ("inconsistent result after apply"). Now the store is authoritative: create seeds it from inline tags, and every Describe* overlays it.

1.6

SubmitJob rejects a non-existent jobQueue (AWS ClientException); DescribeComputeEnvironments reports containerOrchestrationType ("ECS").

Tests

New tags-parity test (create-inline -> ListTags; TagResource -> Describe; containerOrchestrationType). ListJobs default-RUNNING + arrayJobId-selector + no-selector-error assertions. Existing tests updated to create their queue first. cargo test -p fakecloud-batch 18 pass; batch control-plane + CFN e2e green.

---

Summary by cubic

Aligns fake AWS Batch API with real AWS behavior to improve polling accuracy and tag consistency. Fixes ListJobs semantics, unifies tag read/write, validates queues, and adds a missing compute environment field.

• Bug Fixes
  • ListJobs now matches AWS: default to RUNNING when no jobStatus; require exactly one selector (jobQueue/arrayJobId/multiNodeJobId) and support array/multi-node selectors; paginate with maxResults (<=100) + nextToken; return full JobSummary fields.
  • Tags: make the per-ARN tag store authoritative; seed it from create-time tags and overlay on Describe* reads to keep Describe* and ListTagsForResource in sync (fixes Terraform tag drift).
  • SubmitJob rejects non-existent jobQueue with a ClientException.
  • DescribeComputeEnvironments includes containerOrchestrationType: "ECS".

^{Written for commit 6303981. Summary will update on new commits.}