Disconnects slow down replication

[boreq]

Problem:

• during replication the pubs disconnect often
• this is perhaps related to their internal connection management logic and not a problem with scuttlego or planetary
• this is especially problematic during initial replication
• the connection is not reestablished until swift calls ssbDisconnectAll and ssbConnectPeer again

Proposed solution:

• replace ssbDisconnectAll and ssbConnectPeer with a function which sets a list peers that we should stay connected to and lets scuttlego reestablish connections e.g. ssbStayConnectedTo
• this is a dynamic alternative to already existing configuration field "preferred pubs"

[boreq]

I am going to wait with doing anything here as I don't know how much of a problem this is. It is just something I noticed during testing. Definitely not something that should stop us from deploying to test flight. Also this is something that would have been a problem before scuttlego as well.

[boreq]

Related to #301.