Task: Generate the entity part of each component's SQL aggregates

A new entity reaches the database with no hand-edited aggregate. Codegen owns the list of a component's entity files; people own only the hand-written files and the prose around them.

• Codegen emits, per component, a create fragment listing each entity table then its notify trigger, and a drop fragment listing each notify trigger drop then its table drop.
• Each component's hand-written create and drop aggregate includes its fragment once, and no longer lists a generated entity file itself.
• A database recreate succeeds, and the set of tables, functions, triggers and policies it builds is identical to the set before the change.
• Adding an entity to a component needs no aggregate edit, and the drift check fails when a fragment is stale.

(Implementation strategy. Written when work starts; key decisions are distilled into the parent story's * Decisions at close, but the plan itself stays — it is the historical record of what we did.)

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.

From the hand-written SQL census in Find why so much SQL is hand-written and generate what codegen can.

Order inside a component does not depend on soft foreign keys: 88 tables are already created before the parent their insert trigger checks, and every recreate succeeds. What does need order: a table before its notify trigger, the trigger drop before the table drop, functions from earlier components (the root create.sql already orders those), and hand-written files that name a table when they are created (SQL-language functions, views, alter scripts), which must come after the fragment.

First check, before changing any aggregate: recreate a scratch database with one component's aggregate reordered to the planned shape, and compare its catalogue with the current one.

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.