Summary

branched from #103 (now rebased on main; the #103 CLI-command commit is merged, so this PR is just the submitter script + the --date-field extension below)

• Adds submit_storage_tier_workflows.py operator tool that queries STAC in 24h windows over a date range and submits one batch-change-storage-tier webhook payload per window to trigger the eopf-storage-tier-batch-job WorkflowTemplate deployed in https://github.com/EOPF-Explorer/platform-deploy/pull/130 via Argo Events
• Adds unit tests (covering time-window generation, STAC querying, dry-run, error handling, and payload structure)
• Updates operator-tools/README.md with full usage docs and a tool-selection reference table

Update: --date-field (window on registration time)

The submitter originally windowed only on sensing datetime. To drain the S2 backlog by registration time (bulk-converted items were registered long after they were sensed), this adds:

• --date-field {datetime,created,updated} (default datetime, backward compatible). Non-default fields swap the pystac-client datetime= range for a CQL2 between filter on the chosen property — same pattern scripts/query_stac.py uses to harvest by updated.
• --storage-class default flipped STANDARD_IA → STANDARD (STANDARD_IA is unused in the deployment; documented commands stay explicit).
• timeout=30 on the webhook POST (ruff S113, on this script's own code) so a hung endpoint can't stall the per-window loop.
• README gains the registered-window backlog example.

Planned use (#182 follow-up): a one-off prod run over sentinel-2-l2a, windowing on --date-field created for items registered 2025-11-01 → (today − 3 months), moving their S3 objects to STANDARD. Dry-run first, then live via port-forward. The Sensor + batch template are date-agnostic (they consume resolved item_ids), so no cluster-side change is needed for this capability.

Why

Changing S3 storage classes for thousands of items over a multi-month date range is too slow when done sequentially on a local machine. The script fans work out to Argo: one workflow per 24h window, with intra-window parallelism handled by the eopf-storage-tier-batch-job template. Empty windows are skipped automatically. Windowing on created lets us target the registration backlog that a sensing-date window would miss.

Test plan

• uv run --group test pytest tests/unit/test_submit_storage_tier_workflows.py passes (incl. new --date-field / CQL2 / default-storage-class cases)
• uv run --group test pytest tests/unit/test_change_storage_tier_commands.py passes
• Full pre-commit gate green (ruff / ruff-format / mypy / pytest)
• --date-field created builds a CQL2 between on created; default datetime path unchanged (asserted via mocked Client.search kwargs)
• Dry run against staging collection
• Live run (port-forward active) submits correct payload shape and logs action: batch-change-storage-tier with item IDs per window