cluster_traffic: owner can be used to have replication traffic go over the account "owning" the stream, versus this traffic going over the system account (cluster_traffic: system/default).

When pushing an updated JWT to all servers, all these servers would correctly update their cluster_traffic setting.

However, if a server was restarted it would "forget" cluster_traffic: owner was set, and revert back to cluster_traffic: system. This would put this single server to be unable to communicate with the remainder of the cluster. Restarting the rest of the cluster would have them also revert back to cluster_traffic: system, which would allow them to communicate again. But, without respecting the cluster_traffic: owner setting on the account.

This PR fixes that by ensuring cluster_traffic can be updated at the same time as that JetStream is enabled for that particular account upon startup.

Due to this issue, if a server with this fix is deployed, it will not be able to communicate with the other servers that had reverted back to cluster_traffic: system. A clean upgrade path for this would be:

• Temporarily update the account to have cluster_traffic: system. If all servers were restarted they were already on this setting. Any servers that were not yet restarted will now agree on this setting.
• Upgrade all servers to the new server version with this fix.
• Update the account to have cluster_traffic: owner again. It should now be remembered even after a server restart.

Signed-off-by: Maurice van Veen github@mauricevanveen.com