feat(scripts): read-only clone mode — add --skip-source-stage-update to import.py (pull-sequences)

Finding

make pull-sequences (clone mode of scripts/data_transfer/ingestion/platform/import.py) is not read-only against the source API: after a successful clone it PATCHes every cloned sequence's annotation on the source to under_annotation (update_source_annotations_stage(), called at import.py:803-809) to "claim" the sequences for workflow A.

That behavior is correct for the production annotation workflow, but it makes it impossible to hydrate a local API from the remote for testing/debugging without mutating remote state:

• There is no flag to skip the source stage update.
• --dry-run skips the source update but also skips the import itself, so it can't be used for hydration.
• The equivalent flag exists in the pipeline-B pull script (pull_sequence_annotations.py --no-stage-update, read-only export) — but not in import.py.

Proposal

Add a --skip-source-stage-update flag to import.py (mirroring the --no-stage-update precedent) that guards the update_source_annotations_stage() call, so clone mode can run read-only against the source:

uv run python -m scripts.data_transfer.ingestion.platform.import \
  --source-annotation-url https://annotationapi.pyronear.org \
  --url-api-annotation http://localhost:5050 \
  --max-sequences 10 \
  --clone-processing-stage ready_to_annotate \
  --skip-source-stage-update

Optionally expose it through the Makefile (e.g. make pull-sequences READ_ONLY_SOURCE=1) and mention it in the README's local-testing notes.

Start in scripts/data_transfer/ingestion/platform/import.py around lines 803-809 and compare the --no-stage-update precedent in pull_sequence_annotations.py. Run the provided clone command with --skip-source-stage-update and verify sequences are imported locally without calling update_source_annotations_stage() on the source; consider the Makefile and README local-testing notes if included.