[upgrade] endpoints timeout too small

[a-palchikov]

On one occasion, the upgrade failed with on the endpoints phase for DNS but the rolling update of the DNS resources had actually succeeded some time later. Consider bumping the endpoint timeout (currently capped at 5m) as the easiest solution. Ideally, the retry loop should be smarter at detecting transient error points so we can loosen the timeout requirement when we know the loop is legitimately taking long.