Blockfetch can get stuck fetching data from a slow peer

[karknu]

There is a problem where Blockfetch will get stuck fetching data from a far-away peer.
The number of peers we’re fetching data from is controlled by MaxConcurrencyBulkSync, which is set to 1.
If a node starts to use a slow peer (for example because it was listed first in the topology file) it will never move away from it. Any possible requests from new peers will be stomped by MaxConcurrencyBulkSync , even if they could complete faster based on deltaQ estimates.
The only way we switch peers if there is enough network disruption that the outstanding request queue completely drain.