There is no need to use the "collect-then-update" pattern for node
insertion during the SQL migration since if we do have any previously
persisted data for the node and happen to re-run the insertion for that
node, the data will be exactly the same. So we can make use of "On
conflict, no nothing" here too.