Return split db-discovery reports as zip bytes

[jiatolentino]

Summary

get_db_discovery_result_report now returns the raw bytes when the Admin Server splits a large discovery report into a zip, instead of decoding the binary as text (which corrupted the archive).

Background

The Admin Server now auto-splits discovery reports above ~1M rows into a zip of numbered CSV parts (see datamasque DM-3165). The endpoint returns application/zip with an X-DM-Download-Format: zip header in that case.

What changed

• get_db_discovery_result_report(...) return type is now Union[str, bytes]:
  • normal CSV report → str (unchanged),
  • split report → bytes (the zip), detected via the X-DM-Download-Format: zip response header.
• Both forms can be passed straight to start_async_ruleset_generation_from_csv, which already accepts a CSV or a zip of CSV parts.

Testing

• pytest — added a test asserting zip bytes are returned (not decoded) when the header is present.
• ruff, ruff format, mypy clean.

Compatibility

• Backward compatible against an older server: with no X-DM-Download-Format: zip header it returns str, exactly as before.
• Requires the Admin Server DM-3165 change to actually emit a zip.

[carlosfunk]

start_async_ruleset_generation_from_csv hard-codes the filename to ruleset.csv despite the contents. Suggest a small update:

elif isinstance(csv_content, bytes):
    content = BytesIO(csv_content)
    if csv_content[:4] == b"PK\x03\x04":
        filename = "ruleset.zip"

[jiatolentino]

@carlosfunk

start_async_ruleset_generation_from_csv hard-codes the filename to ruleset.csv despite the contents. Suggest a small update:

elif isinstance(csv_content, bytes):
    content = BytesIO(csv_content)
    if csv_content[:4] == b"PK\x03\x04":
        filename = "ruleset.zip"

Fixed. It now detects the zip magic (PK\x03\x04) on bytes input and uploads as ruleset.zip with content_type="application/zip" (kept ruleset.csv/text/csv for everything else). Added a test covering the zip case.