Auto sharding Datalog store

[huahaiy]

To enhance scalability of Datalevin Datalog store, we can distribute data into multiple LMDB files based on hashing of entity id, so data are evenly distributed. Transaction isolation is still managed in the scope of a transaction in a single meta info LMDB, so ACID can be retained.

Because LMDB is B-tree based and has a single writer, as the data volume grows larger, everything get slower and slower. Automatically sharding the data according to data volume also allows concurrent storage as we can write to different shards at the same time. This should improve read/write performance at high data volume.

[huahaiy]

Use this https://en.wikipedia.org/wiki/Rendezvous_hashing