Task: Bring ores.reporting to the clean standard

ores.reporting passes the Component Clean Standard for a component of kind Entity.

• Every checklist item that applies to a component of kind Entity passes, with its evidence recorded on the task.
• Every item that does not apply is recorded as not applicable, with the reason.
• Every generator defect found is fixed at its source, with a regression test.
• ores.reporting is listed in COMPONENTS_UNDER_TEST, and check_component_drift.py --all passes.

The branch is 140 commits behind main, and a merge-base check over the files this pass fixed shows that two of its fixes have been made upstream independently. They are not conflicts to resolve in this branch's favour:

That two of this pass's ten defects were found and fixed upstream while it ran is its own evidence: they were real, they were not local, and the merge is the point at which the duplicate work is discarded rather than carried.

State at the end of this pass, verified: the working tree is clean, `check_component_drift.py –component reporting` exits 0, and all four reporting suites pass.

The objective's second question is answered, twice over. A report triggered from the shell runs the seven-step workflow to `completed`, with the instance travelling `pending`, `running`, `completed` and the FSM guard admitting each transition. It was proven once on the database as it had drifted and once on a fresh `db recreate`, so the result does not depend on accumulated state. Ten defects stood in the way, nine of them outside the reporting component.

The checklist has a recorded result for all 49 items. Passing: B01, B03, B04, B06, G01, G02, G05, G07, M01, M02, M03, M09, M10, H05, P03, P04, V01, V02, V03, V05, V06, V07, W01, W02, W03. Partial and measured: V04 (thirteen commands; three families and eight step commands remain, the latter recorded as unable to answer with the cause diagnosed) and V08 (four of six files have their logic under test). Waiting on ores.dq: M06 and P02's remainder.

What to do next, in order.

1. Rebase. The branch is well behind main and two conflicts already have a disposition: take upstream's version of `workflow_engine.cpp` (`73442db01d` sets `modified_by`, the same defect fixed here) and of `storage_routes.cpp` (`eb3331ca63` deletes the bucket allow-list outright). Re-apply the `reporting` entry to `component_registry.py` on top of whatever landed there.
2. V08's last two files. `report_input_bundle_repository.cpp` needs a database test, following the pattern of the existing eventing integration tests.
3. V04's tail. The three remaining families are template repetition of the one already exercised; the eight step commands cannot answer outside a workflow and need the handler contract changed to reply `bad_request` instead of returning in silence, which is an `ores.service` change and not this component's.
4. Captures worth filing, each with its evidence on this task: the twin gate cannot see a wire type declared outside a `*_protocol.hpp`; nothing checks that a declared subject has a subscriber (twenty did not); the populate scripts cannot be replayed against a populated database because they pass `current_user`; the generated per-entity registrar family exists in twelve components with no caller; and the report instance's `output_message` is overwritten by compensation so it names a step that succeeded.

Work the Component Clean Standard in its section order, one entity at a time, and record each item's result in the table below.

1. Baseline and inventory (B01 to B05). Record the baseline before any edit.
2. Models (M01 to M10).
3. Protocol (P01 to P06).
4. Generation (G01 to G06).
5. Wiring and data (W01 to W04).
6. Shell (S01, S02).
7. Structure and hygiene (H01 to H03).
8. Verification (V01 to V06), ending with the registry entry.

The report_definition bind cannot simply be changed, because the profile catalogue has no exact match and a forced bind is what M02 forbids. Its effective features are tenant, workspace, uuid primary key, change-reason cache and pagination. The three candidates and their mismatches:

So the honest outcome for this entity is a recorded no-match with its actual feature values, which M02 allows, rather than the nearest profile. Deferred to the next unit because the choice changes generated SQL and C++, and it must be made deliberately rather than by trial.

The scheduler-trigger and report-FSM findings live on the sibling task Make the scheduler-triggered report execution correct. That task found the trigger path dead, which is why V03 and V04 of this task cannot pass until it is fixed.

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.

Delivered. The component is listed in COMPONENTS_UNDER_TEST and every gate that reads the registry is green.

Passing: B01, B03, B04, B06, G01, G02, G05, G07, M01, M02, M03, M09, M10, H05, P03, P04, V01, V02, V03, V05, V06, V07, W01, W02, W03.

Both hand-written protocol headers were retired, so the twin gate passes with 45 generated headers across 8 components. The FSM's seeded transitions are enforced by a database trigger for the first time, which exposed that the seed could not represent its own concurrency policy. The registrar now serves all nine verbs per family where it served four, leaving twenty subjects with a generated message, a generated command and no handler. The retired Qt widget column is gone, and removing it changed no generated output at all.

Partial, and measured. V04: thirteen shell commands exercised and answering; the remaining families are template repetition of the one proved, and the eight step commands are recorded as unable to answer with their cause diagnosed. V08: four of six files, with three pure units extracted and tested and the I/O recorded.

Waiting on ores.dq. M06 and P02's remainder, because the surviving hand-written wire type carries a DQ type that DQ has neither modelled nor projected, so there is no TypeScript type to map it to.

Recorded rather than fixed, each with its evidence on the task: the twin gate cannot see a wire type declared outside a *_protocol.hpp; nothing checks that a declared subject has a subscriber; the populate scripts cannot be replayed against a populated database because they pass current_user; the generated per-entity registrar family exists in twelve components with no caller; and compensation overwrites a report's own diagnosis. The handoff section lists the next four steps in order.