How to efficiently get all heads *prior* to a node?

Nobody has claimed this yet.

When building a [hyperdb-index](https://github.com/noffle/hyperdb-index), a common desire is to, when processing a node, to remove all previous heads from the index for that key, and replace them with the new value.

With hyperlog, `node.links` gave those hashes. How could this be done with hyperdb? I hacked it together in a roundabout way:

1. modify the new node's `node.clock` back one entry for its local feed
2. convert `node.clock` into a buffer of heads compatible with `hyperdb#version`
3. do a `hyperdb#checkout` of that version (one entry before this new node was inserted)
4. do a lookup on `node.key`: these are the old heads

This works, but it's a pretty heavy process, especially in the tight loop of an indexer.

@mafintosh can this be done more efficiency with the already existing incremental index?

Start by reading the incremental index behavior and the existing hyperdb#version and hyperdb#checkout entry points described in the issue. Trace how node.clock and node.key are used in the workaround, then determine whether the incremental index can expose the prior heads efficiently. Done means the project has a decided, documented approach or an implemented API for retrieving those heads.

Written by the indexing model from the issue text.