remove redundancy in target extraction from publisher

[mortelli]

As it is, the content recovery hook is working out targets—based on the publisher in the manifest—for each chunk that needs to be repaired.

This is inefficient, as the owner will be determined (plus its feed queried) each time a chunk for the same publisher is attempted to be recovered.

Rather than a feeds handler, the recovery hook should use a struct which gets and sets known targets for any given publisher.

This struct would:

• ~work out the Owner field based on the publisher var~ owner is not publisher
• query the related feed and extract the targets
• register these targets in memory for later re-use

In this manner, the calculations from publisher to targets will be done only once per publisher.

With each subsequent call, the hook would then consult this struct to get the target list and avoid repeated operations & queries.

As a result, the injection of the publisher in the context should be replaced with the injection of the targets instead.

---

Additionally, check if lookup.NoClue could be replaced with something smarter once we know when the last update happened for a feed.