feat(status): раздел «Статус» — контейнеры, карточка, проверки (#14)

- docker-socket-proxy в Compose (только GET /containers/*), сокет только у него
- app/runtime/docker.py: read-only клиент; видны только контейнеры с меткой
  tgit.managed=true, чужие id отклоняются до запроса
- app/runtime/checks.py: running, health, рестарты за окно, CPU/RAM, ошибки
  в логе, HTTP-проба; у каждой проверки свой таймаут, отчёт всегда полный
- app/runtime/snapshots.py: снапшоты раз в 15 с в container_snapshots
- бот: список с 🟢🟡🔴⚪, карточка, «🩺 Запустить проверки», «🔄 Обновить»
- деплой ставит метки tgit.sha / tgit.deployed_at на контейнер бота
- имя джоба CD без неотрисованного выражения

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

app/runtime/snapshots.py

@@ -0,0 +1,87 @@
+"""Every 15 s: restart count, CPU and RAM of each running managed container → `container_snapshots`.
+
+The history answers «are restarts growing over the last 5 minutes» (S-3) and feeds alerts (#15).
+Old rows are removed by `app.db.maintenance`.
+"""
+
+import asyncio
+import logging
+from datetime import datetime
+
+from sqlalchemy import select
+
+from app.db.models import ContainerSnapshot, utcnow
+from app.db.session import SessionMaker
+from app.runtime.docker import DockerClient, DockerError
+
+log = logging.getLogger(__name__)
+
+
+class SnapshotStore:
+    def __init__(self, sessions: SessionMaker) -> None:
+        self._sessions = sessions
+
+    async def add(self, rows: list[ContainerSnapshot]) -> None:
+        if rows:
+            async with self._sessions.begin() as session:
+                session.add_all(rows)
+
+    async def oldest_since(self, container: str, since: datetime) -> ContainerSnapshot | None:
+        async with self._sessions() as session:
+            return await session.scalar(
+                select(ContainerSnapshot)
+                .where(ContainerSnapshot.container == container, ContainerSnapshot.at >= since)
+                .order_by(ContainerSnapshot.at)
+                .limit(1)
+            )
+
+    async def latest(self, container: str) -> ContainerSnapshot | None:
+        async with self._sessions() as session:
+            return await session.scalar(
+                select(ContainerSnapshot)
+                .where(ContainerSnapshot.container == container)
+                .order_by(ContainerSnapshot.at.desc())
+                .limit(1)
+            )
+
+
+async def _snapshot(docker: DockerClient, container_id: str, at: datetime) -> ContainerSnapshot | None:
+    try:
+        details, stats = await asyncio.gather(docker.inspect(container_id), docker.stats(container_id))
+    except DockerError as e:
+        log.warning("snapshot %s: %s", container_id[:12], e)
+        return None
+    return ContainerSnapshot(
+        container=details.name,
+        at=at,
+        restart_count=details.restart_count,
+        cpu_pct=stats.cpu_pct,
+        mem_pct=round(stats.mem_pct, 1),
+        health=details.health,
+    )
+
+
+async def collect_once(docker: DockerClient, store: SnapshotStore, now: datetime | None = None) -> int:
+    at = now or utcnow()
+    running = [c for c in await docker.containers() if c.state == "running"]
+    rows = await asyncio.gather(*(_snapshot(docker, c.id, at) for c in running))
+    snapshots = [r for r in rows if r is not None]
+    await store.add(snapshots)
+    return len(snapshots)
+
+
+async def snapshot_loop(docker: DockerClient, store: SnapshotStore, interval: float) -> None:
+    failing = False
+    while True:
+        try:
+            await collect_once(docker, store)
+            if failing:
+                log.info("snapshots: docker reachable again")
+            failing = False
+        except DockerError as e:
+            if not failing:  # one line per outage, not one every 15 s
+                log.warning("snapshots: %s", e)
+            failing = True
+        except Exception:  # a broken round must not kill the bot
+            log.exception("snapshots failed")
+        await asyncio.sleep(interval)