Prolly tree projected onto key-value storage

• [_] node
• [_] tree
• [_] iterator
• [_] encoder

TODO

• currently we read all values on level0, but that's problematic:
  • if we store several trees / generations, all of them will be read
  • need to store root for each generation: root:generation
  • then Deserialize starting from that root and read only the relevant level0 nodes
    • problem: the tree currently doesn't store children hashes in the nodes
    • we store hashes and keys (boundaries), but if we store multiple generations, that's not enough
• switch key from int to string so that I can experiment on user library
• build trees for the latest N generations of a library
  • sample data
  • see how storage size grows vs naive approach (saving JSONs) -- sample data
  • igor's data
  • igor's data growth

• partition storage by adding prefix, e.g. based on generation / timestamp
• make a db to index changes between generations
• kv iterator so that I can use it in Diff, compare on KV level without loading the whole tree