#3251 autoGc does not enforce version caps for scheduled workflow steps

Summary

With autoGc: true in .swamp.yaml, model resources written by a scheduled workflow step continue accumulating beyond their declared garbageCollection version caps. Explicit swamp data gc correctly restores the caps.

This affects long-running swamp serve deployments with frequent schedules. The fix likely needs the repository autoGc setting to be propagated into model-method execution initiated by the workflow scheduler, using the same automatic GC path as direct model-method execution.

Reproduction

1. Use the filesystem datastore and set autoGc: true in .swamp.yaml.
2. Define a model with resources using integer garbageCollection caps. In the observed case, three resources had caps of 100, 30, and 100.
3. Create a scheduled workflow that invokes the model method once per minute and writes new versions of those resources.
4. Run swamp data gc --force --json to establish a clean capped baseline.
5. Allow 16 scheduled executions to complete successfully.
6. Inspect each artifact with swamp data versions <model> <artifact> --json.

Observed result

The three resources contained 116, 46, and 116 versions respectively: each was 16 versions above its configured cap. A subsequent swamp data gc --dry-run --json reported excess versions eligible for deletion.

Expected result

When autoGc: true, write-time or post-method automatic GC should keep scheduled workflow outputs within their declared version caps without requiring a separate manual GC command.

Environment

• Swamp tested: 20261011.153055.0-sha.8453c3cc
• Latest source checked: 20261011.154850.0-sha.d650cb70
• Execution mode: swamp serve, local filesystem datastore, cron-scheduled workflow
• Platform: Linux with systemd

Additional evidence

Inspection of the latest source found autoGc read from the repository marker and passed into direct model-method execution. The workflow/serve execution path did not show corresponding propagation of that setting.

Workaround

A systemd timer periodically runs both swamp data gc --force --json and swamp run gc --force --json.