fix(api): refresh drift snapshot and preserve AWS backup external IDs

crates/clickhouse-cloud-api/README.md

@@ -72,6 +72,8 @@ The beta `replica_aware_routing_get` and `replica_aware_routing_update` service
 
 The MySQL and Postgres ClickPipe source authentication enums (`ClickPipeMutateMySQLSourceAuthentication`, `ClickPipeMySQLSourceAuthentication`, `ClickPipePatchMySQLSourceAuthentication`, `ClickPipeMutatePostgresSourceAuthentication`, `ClickPipePostgresSourceAuthentication`) gain `ServiceAccountWorkloadIdentity` (`SERVICE_ACCOUNT_WORKLOAD_IDENTITY`), and the MySQL source type enums gain `Cloudsqlmysql` (`cloudsqlmysql`); exhaustive `match`es over them need the new arm. `ClickPipeMongoDBPipeTableMapping` gains `structured_ingestion` (`ClickPipeMongoDBStructuredIngestion`: required `enabled`, plus optional `columns` of `ClickPipeMongoDBPipeColumn` `name`/`type` pairs, `drop_unexpected_values`, `excluded_columns`, `sorting_keys`, and `partition_by_expr`) to replicate a collection into typed columns instead of one JSON column; responses use the all-`Option` `ClickPipeMongoDBStructuredIngestionResponse` and `ClickPipeMongoDBPipeColumnResponse`. Struct-literal callers of the table mapping need `structured_ingestion: None` or `..Default::default()`.
 
+AWS backup-bucket models now carry `iam_role_external_id` (`iamRoleExternalId`). `AwsBackupBucket` responses tolerate missing and null IDs. The current spec requires a string on `AwsBackupBucketPostRequestV1` and `AwsBackupBucketProperties`; supply it when constructing or deserializing these request types. `AwsBackupBucketPatchRequestV1.iam_role_external_id` distinguishes omission (`None`, preserve), a value (`Some(Some(id))`, replace), and explicit null (`Some(None)`, clear). Its deserializer preserves all three states. Regenerated operation metadata also adds `control-plane:service:manage` to `INSTANCE_CREATE`, alongside `control-plane:organization:create-service`.
+
 The Query API endpoint management methods are `query_api_endpoint_create`, `query_api_endpoint_get`, `query_api_endpoint_list`, `query_api_endpoint_update`, and `query_api_endpoint_delete`. Create and update take `PublicQueryApiEndpointRequest`; list accepts an optional cursor and limit (1–100) and returns `ApiResponse<Vec<PublicQueryApiEndpointListItem>>`. Read entries from `result` and pagination from the envelope's `next_cursor`, `limit`, and `total_count`. The earlier `QueryApiEndpointListResponse` container remains available for existing callers, but is no longer the list method's return payload. User-owned endpoints can be listed and read, but cannot be updated or deleted through this API.
 
 The saved query methods (no longer beta) are `saved_query_create`, `saved_query_get`, `saved_query_list`, `saved_query_update`, and `saved_query_delete`, scoped to a service. Create and update take `PublicSavedQueryRequest` (update replaces the whole query); get returns `PublicSavedQuery` and list returns `PublicSavedQueryListItem` entries. List accepts an optional cursor and limit (1–100) and reports pagination on the envelope: `ApiResponse::next_cursor` (`None` on the last page), `limit`, and `total_count`.