Refine merkle tree construction

[tbekas]

Resolves #511

[tbekas]

Adressing the comment by @elcritch from previous PR.

I added some docs to individual methods instead of a module.

[elcritch]

Also I created an issue for the pre-image attack bit: #511. You could probably close it with this PR?

[tbekas]

Also I created an issue for the pre-image attack bit: #511. You could probably close it with this PR?

I just marked this PR as resolving the #511

[dryajov]

From our sync discussion:

• We want to store the global multihash codec for the interior nodes as a type field and only store the digest bytes for each node
• We want to additionally store the CID codec for the merkle root so we can continue using the existing repo as a storage backend
• We should probably rework the tree to be a full tree, which means that it has to contain an even number of nodes at each level
  • The overhead from that would be one additional hash per level, but it will greatly simplify the logic for calculating the number of nodes, height, offsets, etc. For reference look at the https://en.wikipedia.org/wiki/Binary_tree#Properties_of_binary_trees
  • For serialization purposes, we don't need to store the duplicate/dummy nodes, so the overhead is mostly just in memory

[tbekas]

@dryajov Let me know if the current data model as well as accessors are sufficient. Internally tree is stored just a seq[byte] while on the API MultiHash is used (raw bytes are wrapped into a MultiHash)

MerkleTree* = object
    mcodec: MultiCodec
    digestSize: Natural
    leavesCount: Natural
    nodesBuffer: seq[byte]
    
proc root*(self: MerkleTree): MultiHash

proc leaves*(self: MerkleTree): seq[MultiHash]

[tbekas]

Moved to a different PR