refactor: remove data_sources_expanded view (#845)

[maxachis]

Summary

Closes #845.

Drops the data_sources_expanded SQL view (and its DataSourceExpanded SQLAlchemy shadow) so future data_sources schema changes no longer require dropping/recreating a view in every migration.

The view's only added value over the data_sources table was the joined record_types.name column. That is now exposed via a SQLAlchemy column_property on DataSource that issues a correlated subquery against record_types, producing the same record_type_name value with the same LEFT JOIN semantics (NULL when record_type_id is unset).

Replacements by call site

• db/models/implementations/core/data_source/core.py — added agencies and data_requests relationships (moved off the deleted DataSourceExpanded) and the record_type_name column_property.
• db/client/core.py, db/dynamic_query_constructor.py, endpoints/instantiations/data_sources_/get/{by_id,by_id/agencies,many}/query.py, endpoints/instantiations/data_sources_/get/convert.py — all select(DataSourceExpanded) / selectinload(DataSourceExpanded.*) switched to the DataSource model. No query semantics change; the column_property injects the same LEFT JOIN record_types projection.
• db/subquery_logic.py, middleware/primary_resource_logic/{data_sources.py,data_requests_/related/related_source/get.py} — Relations.DATA_SOURCES_EXPANDED references repointed to Relations.DATA_SOURCES.
• db/models/implementations/core/{agency/core.py,data_request/expanded.py} — data_sources relationships now point at DataSource instead of DataSourceExpanded.
• db/models/table_reference.py, middleware/enums.py, middleware/column_permission/mapping.py — removed data_sources_expanded registrations; merged record_type_name into the data_sources permission entry (same role/permission settings as the deleted data_sources_expanded entry).
• tests/utilities/test_result_formatter.py — instantiates DataSource directly.
• New migration alembic/versions/2026_05_02_0839-a3fcd7a36a70_drop_data_sources_expanded_view.py drops the view in upgrade() and recreates it in downgrade() for reversibility.

Test plan

• uv run ruff check . — clean
• uv run basedpyright --level error — 0 errors
• uv run alembic upgrade head then downgrade -1 then upgrade head against a local PG15 — view drops and re-creates as expected
• uv run pytest tests/ (all sub-suites) — 257 passed locally; same tests_comprehensive failures (11 fail / 7 error) exist on dev baseline and are unrelated
• CI run_pytest.yml against full DB

Risk areas for reviewer

• The column_property runs a correlated scalar subquery for record_type_name whenever it is loaded. For the existing call sites this matches the LEFT JOIN the view used. If any caller did data_sources_expanded.record_type_name filtering in a WHERE clause it would still work (SQLAlchemy will inline the subquery), but no such filter exists in the current codebase — worth a quick second look.
• The DataSourceExpandedDTO / DataSourceExpandedSchema names were left in place since they describe the API output schema, not the database view. Renaming is out of scope.

🤖 Generated with Claude Code