feat: file and link counts on opportunities and notices (API 5.9.0)

[makegov-mark]

What

Re-vendors contracts/filter_shape_contract.json at Tango API 5.9.0 and regenerates tango/shapes/generated_overlay.py.

Tango API 5.9.0 splits attachment counts by attachment type:

• Opportunities: meta.files_count and meta.links_count, beside the existing meta.attachments_count.
• Notices: file_count and link_count, beside the existing attachment_count.
• Vehicles: the embedded opportunity accepts opportunity(meta(files_count,links_count)).

A file is a document attached to the notice, and only files have extracted text.
A link is a URL the notice lists.
The existing totals still count every attachment, links included, and an attachment of any other type counts only in the total, so files plus links need not equal it.
The new fields are None until a record has been counted, so None means unknown, not zero.

Contract changes, 5.8.0 to 5.9.0

• notices: shape fields file_count, link_count added.
• opportunities: shape fields meta.files_count, meta.links_count added.
• vehicles: shape fields opportunity.meta.files_count, opportunity.meta.links_count added.
• entities: the state and zip_code filter entries no longer carry a lookup annotation. The filters themselves, their names and their types are unchanged, and the SDK's gates do not read that annotation.
• meta.api_version: 5.8.0 to 5.9.0.

No filter, ordering field, endpoint or resource was added or removed, so no client method changed.

Changes

• Generated: the vendored contract and four new FieldSchema entries in the shape overlay (all int, optional); the opportunity embedded in a vehicle shares the opportunity meta schema, so it needed no entry of its own.
• By hand: Notice gains attachment_count, file_count and link_count attributes; the Opportunity and Notice docstrings and docs/API_REFERENCE.md describe the counts and the None caveat; tests; changelog.
• No default SDK shape changed.

Testing

• uv run python scripts/check_filter_shape_conformance.py: 0 errors, 24 warnings (the same 24 as against the 5.8.0 contract).
• uv run python scripts/check_shape_coverage.py: 14 gaps, 14 baselined, 0 new. Before regenerating it reported the 6 new fields above.
• uv run pytest: 609 passed, 31 skipped.
• uv run ruff format --check tango/, uv run ruff check tango/, uv run mypy tango/: clean.
• The new tests fail against the previous overlay and pass against the regenerated one.
• Cassettes are untouched: no existing integration test requests these fields.

Release

Not bumped here.
The changelog entry sits under [Unreleased] for the next release PR.

🤖 Generated with Claude Code