[FEATURE] chunkserver to store small chunks in chunk-volumes

[onlyjob]

Chunkservers can greatly optimise performance with many small files by aggregating small chunks into chunk-volumes. The idea is to implement a special storage backend in chunkserver code that stores small chunks in 64 MiB volume-files to reduce overall number of chunk-files. Implementation can leverage LevelDB, SQLite3 or a similar technology for a backend.

Chunkserver's performance suffer dramatically when chunkservers handle millions of files, which is a case when MooseFS is used to store Maildir structures, git repositories and CCTV data. SeaweedFS implementation stores all chunks in volume files (not a chunk-per-file like MooseFS) -- that achieves superior performance on great number of small files.

My original feature request #370 was moved to discussion for some reason so I'm filing feature request again, this time described more precisely. Thanks.