Currently buckets namespaces can be created as normal collections (non-timeseries), but cannot be sharded. This is causing issue for PM-3364 project as the coordinator needs to support bucket namespace in case of unsplittable collection creation. Before, the coordinator could simply assume a collection already existed as timeseries if a bucket namespace was found on the CSRS. This assumption is no longer true. This ticket is to create a couple of tests where all the combination of create/shard collection in combination with already existing "bucket" namespace to know already what to expect. 

TODO create table