Add pytest coverage for automation/run_automation.py

[katherine-stansifer]

Context

automation/run_automation.py is the head-container entrypoint that the startOntBasecall Lambda submits to AWS Batch. It currently has no tests — its argparse surface, the Nextflow argv builder, and the in-process seq_import.samplesheet.generate_samplesheet call are all only exercised end-to-end against real deliveries.

Came out of PR #23 review (the refactor from subprocess.run("python -m seq_import …") to a direct import made the lack of test coverage more visible).

Scope

Add a tests/ directory with pytest coverage for automation/run_automation.py:

• parse_args — required flags reject missing input; dashed flags map to expected attribute names (--aws-queue → args.aws_queue, etc.).
• build_nextflow_cmd — golden-test the argv given a representative (delivery, kit, aws_queue, base_bucket, work_bucket) tuple. Covers the s3://… URL composition and the -profile batch / --barcodes …/supplemental/barcodes.tsv conventions.
• main — mock subprocess.run and seq_import.samplesheet.generate_samplesheet (plus boto3.client) and assert: nextflow runs first with check=True and cwd=/workflow; samplesheet generation runs only on nextflow success; the delivery + base-bucket flow through to generate_samplesheet as delivery= / bucket=.

Out of scope

• Integration tests that actually invoke Nextflow or hit S3 — the unit coverage above is the floor; a heavier integration harness can come later if we need it.

Wiring

• Add pytest to automation/environment.yml (dev-only is fine; we already ship the env into the image).
• Add a GitHub Actions job to .github/workflows/ that runs pytest on PRs. The existing docker-build.yml is a reasonable template.