Add sink-level dequeue retries (finally)

[jakthom]

If batches fail to send to a downstream sink, they should retry up to N times with an exponential backoff.

Segment has some nice In Production stats here:

https://segment.com/blog/introducing-centrifuge/

Specifically:

On average, we find about 1.5% of all global data succeeds on a retry, where it did not succeed on the first delivery attempt.

and :

On the graph below, you can see all successful retries split by ‘attempt number’. We typically deliver the majority of messages on their second try (the large yellow bar), but about 50% of retries succeed only on the third through the tenth attempts. 

[jakthom]

Some things immediately come to mind here:

• Could the centrifuge model be used directly, but using sqlite?
• What happens when failures exhaust retries?
• Will saturating a sink-specific buffered channel back up the whole pipeline if one sink fails? (Yes probably). How can this be avoided?

[jakthom]

Completed in https://github.com/silverton-io/buz/pull/565

Went the simple route for now - additional mechanisms can be provided @ manifold level.