feat(chain)!: `O(n)` canonicalization algorithm

feat(chain)!: `O(n)` canonicalization algorithm

This is an O(n) algorithm to determine the canonical set of txids.

* Run 1: Iterate txs with anchors, starting from highest anchor height
  txs.
* Run 2: Iterate txs with last-seen values, starting from highest
  last-seen values.
* Run 3: Iterate txs that are remaining from run 1 which are not
  anchored in the best chain.

Since all transitively-anchored txs are added to the `canonical` set in
run 1, and anything that conflicts to anchored txs are already added to
`not_canonial`, we can guarantee that run 2 will not traverse anything
that directly or indirectly conflicts anything that is anchored.

Run 3 is needed in case a tx does not have a last-seen value, but is
seen in a conflicting chain.

`TxGraph` is updated to include indexes `txids_by_anchor_height` and
`txids_by_last_seen`. These are populated by the `insert_anchor` and
`insert_seen_at` methods. Generic constaints needed to be tightened as
these methods need to be aware of the anchor height to create
`LastSeenIn`.