fix(backup): reimport the postgres dump on restore (restore was a no-op)

The recipe dumped the DB on backup but shipped NO backupbot.restore.post-hook, and
archived the whole live PGDATA dir — so a restore extracted files under the running
postgres without reloading them, silently keeping the un-restored (live) state. A
restored backup therefore lost all data written since the snapshot.

Switch to the coop-cloud /pg_backup.sh convention (as matrix-synapse): backup =
pg_dump|gzip -> backup.sql; restore = terminate connections, FORCE-drop, recreate,
reimport the dump deterministically. Archive just backup.sql, not the whole PGDATA dir.

compose.yml

@@ -58,9 +58,13 @@ services:
     deploy:
       labels:
         backupbot.backup: "true"
-        backupbot.backup.pre-hook: "PGPASSWORD=$$(cat $${POSTGRES_PASSWORD_FILE}) pg_dump -U $${POSTGRES_USER} $${POSTGRES_DB} > /var/lib/postgresql/data/postgres-backup.sql"
-        backupbot.backup.post-hook: "rm -rf /var/lib/postgresql/data/postgres-backup.sql"
-        backupbot.backup.path: "/var/lib/postgresql/data/"
+        backupbot.backup.pre-hook: "/pg_backup.sh backup"
+        backupbot.backup.volumes.postgres_data.path: "backup.sql"
+        backupbot.restore.post-hook: "/pg_backup.sh restore"
+    configs:
+      - source: pg_backup
+        target: /pg_backup.sh
+        mode: 0555
 
 
 secrets:
@@ -69,6 +73,9 @@ secrets:
     name: ${STACK_NAME}_postgres_password_${SECRET_POSTGRES_PASSWORD_VERSION}
 
 configs:
+  pg_backup:
+    name: ${STACK_NAME}_pg_backup_${PG_BACKUP_VERSION}
+    file: pg_backup.sh
   abra_mattermost_entrypoint:
     name: ${STACK_NAME}_entrypoint_${ABRA_MATTERMOST_ENTRYPOINT_VERSION}
     file: ./entrypoint.sh