Optimizes reading multiple chunks from a shard.
Serial calls to .get() in the sharding codec have been replaced with
a single call to .get_partial_values() which stores may optimize by making
concurrent requests and/or coalescing nearby requests to the same shard.