feat(backup): stricter databases.csv semantics + atomic SQL dumps

- read databases.csv with stable types (dtype=str, keep_default_na=False)
- validate database field: require '*' or concrete name (no empty/NaN)
- support Postgres cluster dumps via '*' entries (pg_dumpall)
- write SQL dumps atomically to avoid partial/empty files
- early-skip fully ignored volumes before creating backup directories
- update seed CLI to enforce new contract and update by (instance,database)
- adjust tests: sql dir naming + add E2E coverage for early-skip and '*' seeding

tests/e2e/test_e2e_dump_only_sql_mixed_run.py

@@ -145,7 +145,7 @@ class TestE2EDumpOnlySqlMixedRun(unittest.TestCase):
     def test_db_volume_has_dump_and_no_files_dir(self) -> None:
         base = backup_path(self.backups_dir, self.repo_name, self.version, self.db_volume)
 
-        dumps = base / "dumps"
+        dumps = base / "sql"
         files = base / "files"
 
         self.assertTrue(dumps.exists(), f"Expected dumps dir for DB volume at: {dumps}")