Stream the public dump's NDJSON files into the archive #192

Problem

The public dump builds each score set's NDJSON files as one string in memory. va_ndjson and vrs_ndjson in scripts/export_public_data.py collect every line into a list, join it and return a str, which is then written into the archive. Peak memory grows with the largest score set in the dump. The api's own NDJSON routes already stream: see _stream_score_set_variant_details in routers/score_sets.py.

Scope

• Make va_ndjson and vrs_ndjson generators that yield lines in batches.
• Write each file into the archive through ZipFile.open(name, "w"), one batch at a time, instead of writestr with a full string.
• Leave the JSON metadata files as they are. They're small.

Acceptance criteria

• The archive's bytes are unchanged for the existing tests/scripts/test_export_public_data.py fixtures.
• Peak memory while exporting the largest score set doesn't grow with its variant count (spot-checked).