Provide a way for clients to pre-split a key range into shards

FDB should provide a way for clients to pre-split a key range into shards based on prior knowledge. For example if the client knew the workload they were about to run would create a very write hot load on a known key range, they can use this mechanism to pre-split that key range into properly sized shards to avoid later data movement caused by DD's write hot detection and splitting logic.

This feature could also be helpful for the bulk loading feature described in #1002