feat: Compute edges from tags rather than storing (#500)

* feat: Compute edges from tags rather than storing

In dense graphs, there may be `O(node^2)` edges for tags with high
connectivity. Storing the tags alone (`O(tags)`) allows for faster
writes. We can further address dense graphs by using queries with
similarity bounds and filters.

Other changes:

- refactor: Introduce `Link` as interface for `LinkTag`
  This allows non-tag based links to be added in the future.
- refactor: remove unused code

Future improvements
- Further performance tuning of queries, including denormalization to
  allow for fewer queries during traversal.
- Consider / experiment with single key queries rather than `IN`.

Author: bjchambers

libs/knowledge-store/README.md

@@ -15,7 +15,7 @@ The Graph Store makes use of the following metadata fields on each `Document`:
 - `content_id`: If assigned, this specifies the unique ID of the `Document`.
   If not assigned, one will be generated.
   This should be set if you may re-ingest the same document so that it is overwritten rather than being duplicated.
-- `link_tags`: A set of `LinkTag`s indicating how this node should be linked to other nodes.
+- `links`: A set of `Link`s indicating how this node should be linked to other nodes.
 
 #### Hyperlinks