connmgr: Only mark persistent peer reconn pending.

[davecgh]

The current code is incorrectly treating non-persistent peer requests as though they will be reconnected to whenever the target number of outbound peers has not been reached by marking them as pending and logging the reconnection attempt that will never come. In addition to the spurious logging, it also means that manually attempting to reconnect to such a peer will incorrectly believe a connection is already pending.

This resolves that issue by only logging the attempt and adding the connection request back to the pending map for persistent peers.