Task: Move the duplicated DQ models onto their owners

Each of the six artefact tables is generated from its owning model, and no duplicate ores.dq model or DQ stack remains.

• The six ores.dq models and their generated stacks are deleted, after checking nothing still consumes them.
• Each owning model carries the flag; the populates, publish functions and artefact types use the regenerated tables.
• The datasets publish as before: pgTAP and an Acme provision show the same rows.

1. Confirm the inventory. Five of the six tables come from a duplicate ores.dq model. risk_report_configs does not: it already comes from ores.reporting.risk_report_config, which carries the flag. The guide's table was wrong about it.
2. For each of the five, set the flag on the owning model. Where the staging shape differs from the entity, write an * Artefact columns section, and carry the DQ model's * Artefact indexes over. A script reads the current artefact table, so the section states the same columns, types, nullability and defaults.
3. Regenerate the SQL schema for each owning model and diff the artefact table against the old one. The table must not change, other than its comment header.
4. Delete each ores.dq model and its generated stack: the DQ API domain, JSON and table I/O, generator and protocol; the DQ core handler, registrar, entity, mapper, repository and service; and the generated TypeScript. Remove the registrar wiring, the shell command, the recipes, the protocol rows and the read permissions only these handlers checked.
5. Follow the provider table rename through the hand-written SQL.
6. Run the drift checks, build, ctest and pgTAP. Recreate the database and provision Acme.

Every number this task claims about a tree, with the command that produced it and the commit it was measured at. A number without both is not a measurement, because a clean-up record describes a tree and trunk moves under a branch: a consumer count, a gate's component count and a suite's size all change with no code changing, and the reader of a bare number cannot tell which tree it came from. Where the raw output matters, put it in an evidence/ file beside the task and name that file in the Command column.

• The provider table is renamed. The owning model's plural is market_data_providers, so the staging table is now ores_dq_market_data_providers_artefact_tbl. The populate, the publish function, the artefact type row, the drop script and the pgTAP test follow the rename. The dataset code and the artefact type code do not change.
• The report definition tier columns move into the model. The tier and the applicable jurisdiction were added by a hand-written alter script, dq_report_definitions_artefact_tier_alter.sql. They are now the last two artefact columns of ores.reporting.report_definition, and the alter script is deleted.
• The report definition template handler stays. dq.v1.report_definition_templates.list reads the artefact table in SQL, not through the deleted repository, and the reporting service calls it. Its permission, dq::report_definitions:read, stays with it.
• The DQ report definition capture is resolved. The capture BAFCC111 said the DQ report definition entity names a table that no script creates. The entity is gone, so the capture is filed with a resolution.
• The CSA, netting agreement and netting set tables keep their natural-key shape: a netting set code, an agreement number and a counterparty LEI where the store holds uuids. Publishing resolves them in the target tenant.

Manual QA scenarios (scaffolded via compass add test_scenario) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

The five duplicated DQ models are gone. csas, netting_agreements and netting_sets come from their refdata models, report_definitions from ores.reporting.report_definition, and the provider staging table from ores.synthetic.market_data_provider, now named ores_dq_market_data_providers_artefact_tbl. Each owning model carries the artefact flag and an * Artefact columns section where the staging shape differs, and the regenerated tables keep their columns. The report definition tier columns are in the model, and the alter script is deleted. The DQ API, core, shell and TypeScript stacks of the five models are deleted, with their registrar wiring, shell command, recipes, protocol rows and permissions.

risk_report_configs needed no change: it already came from its reporting model, and the DQ guide now says so.

Acceptance: the models and stacks are deleted, and nothing consumed them other than the wiring removed here. The owning models carry the flag. pgTAP, ctest and an Acme provision pass, and the datasets publish into the Acme tenant. The provision recorded one failed step, start_market_feeds, which timed out while the feeds started. This branch does not touch that path; the capture start_market_feeds_step_times_out_while_the_feeds_start records it.