[Update] Document behavior of DROP DATABASE on multi-node

[erimatnor]

When dropping a distributed database, the corresponding databases are not dropped on the data nodes. This is to prevent accidentally dropping data that should be preserved on those nodes. Further, since the databases could have existed prior to incorporating them into a multi-node cluster, they might have had data prior to being part of a distributed database.

However, keeping the database on a data node instance prevents adding them back. The databases either have to be dropped manually on each data node or the dist_uuid removed in the metadata on each data node DB.

Until we have a more seamless solution for how to handle this, we should document the current behavior and describe how to "add back" a data node using the same database.