As part of WT-8366, we realised that __wt_txn_publish_durable_timestamp is somewhat mis-named. Alex G summed it up best:

"It assumes that everyone knows that a commit timestamp implicitly becomes the durable timestamp if a different one isn't set. I think it could be renamed to wt_txn_publish_user_timestamp, and its intent would be clearer. It publishes any timestamp for a transaction that needs to be published."

We should check if (a) this is what we really want the function to do, and (b) rename it accordingly.