feat(dq): добавлен контроль качества конвейера

- Зачем:
  - конвейеру нужна публикуемая сводка текущего качества с красным завершением при нарушении договоров.
- Что:
  - добавлены DDL, публичное представление и пять сверок на естественном деловом зерне.
  - даг dq_check встроен после DM и публикует полные дневные партиции до именованных утверждений.
  - параллелизм DDS, DM и DQ приведён к измеренному лимиту памяти планировщика.
- Проверка:
  - make lint; make config-test; make rebuild-storage.
  - make smoke; make check-clickhouse; make check-services.
  - три семейства проверок покрашены искусственными расхождениями и восстановлены повторным полным прогоном.

sql/dq/purchase_vs_orders_vs_dds.sql

@@ -0,0 +1,95 @@
+-- DQ: строка сверки покупок и заказов против независимой сборки из DDS.
+--
+-- Проверяются все деловые поля. Счёт физических строк отдельно ловит дубль
+-- ключа, даже если обе копии несут одинаковые значения.
+WITH
+    purchases AS
+    (
+        SELECT
+            purchase_id[1] AS purchase_order_id,
+            any(event_date) AS purchase_day,
+            any(purchase_revenue[1]) AS declared_revenue,
+            count() AS purchase_count
+        FROM dds.event_v
+        WHERE event_type = 'purchase'
+        GROUP BY purchase_order_id
+    ),
+    orders AS
+    (
+        SELECT
+            order_id AS backend_order_id,
+            order_date,
+            items_total,
+            status
+        FROM dds.order_v
+    ),
+    reference AS
+    (
+        SELECT
+            coalesce(order_date, purchase_day) AS order_day,
+            coalesce(purchase_order_id, backend_order_id) AS order_id,
+            declared_revenue,
+            items_total,
+            status,
+            multiIf(
+                backend_order_id IS NULL, 'awaiting_order',
+                status = 'cancelled', 'cancelled',
+                purchase_order_id IS NULL, 'lost_event',
+                purchase_count > 1, 'duplicate_event',
+                declared_revenue != items_total, 'amount_delta',
+                'match'
+            ) AS mismatch_class
+        FROM purchases
+        GLOBAL FULL OUTER JOIN orders
+            ON purchase_order_id = backend_order_id
+    ),
+    actual AS
+    (
+        SELECT
+            order_day,
+            order_id,
+            count() AS physical_rows,
+            any(declared_revenue) AS declared_revenue,
+            any(items_total) AS items_total,
+            any(status) AS status,
+            any(mismatch_class) AS mismatch_class
+        FROM dm.purchase_vs_orders_v
+        GROUP BY
+            order_day,
+            order_id
+    )
+SELECT
+    coalesce(reference.order_day, actual.order_day) AS data_date,
+    coalesce(reference.order_id, actual.order_id) AS business_key,
+    isNotNull(reference.order_id) AS reference_present,
+    isNotNull(actual.order_id) AS actual_present,
+    toString(tuple(
+        reference.declared_revenue,
+        reference.items_total,
+        reference.status,
+        reference.mismatch_class
+    )) AS reference_value,
+    concat(
+        toString(tuple(
+            actual.declared_revenue,
+            actual.items_total,
+            actual.status,
+            actual.mismatch_class
+        )),
+        ', rows=',
+        toString(actual.physical_rows)
+    ) AS actual_value,
+    isNull(reference.order_id)
+        OR isNull(actual.order_id)
+        OR actual.physical_rows != 1
+        OR NOT isNotDistinctFrom(reference.declared_revenue, actual.declared_revenue)
+        OR NOT isNotDistinctFrom(reference.items_total, actual.items_total)
+        OR NOT isNotDistinctFrom(reference.status, actual.status)
+        OR NOT isNotDistinctFrom(reference.mismatch_class, actual.mismatch_class)
+        AS failed
+FROM reference
+GLOBAL FULL OUTER JOIN actual USING (order_day, order_id)
+ORDER BY
+    data_date,
+    business_key
+SETTINGS join_use_nulls = 1