Separate topology discovery/generation from server start-up

To address gateway start-up performance, it may be good to skip discovery at that time IFF cluster configuration monitoring is enabled.

This way, existing topologies can be quickly deployed as-is during start-up, and then the config monitor can drive any necessary changes more discretely, re-generating only those affected topologies if any at all.

Whether the condition of requiring config monitoring enabled for this is correct is up for discussion.