Pool ns_records update crashes when tenant has NS records

Pool update will fail on _delete_ns if tenants have their own non-managed NS recordset

Stack trace is quite misleading, as the _delete_ns code also makes the assumption that there will only be one NS recordset, and crashes before the actual issue (get_recordset vs get_recordsets).

Bug is present on all current designate versions.

Patch is attached which applies to stable/stein and will apply to master with some context adjustment, the relevant code has not changed.
Patch adds querying+filtering identical to what is done in _add_ns, and works as intended in my cloud.