USN publishing process has a race condition

[Steve Beattie]

When we prepare USN updates, we run UCT/scripts/sis-generate-usn to generate a script for the USN. That scripts pushes information to the primary USN database on lillypilly.c.c. If someone is preparing an update to the point of running the sis-gererate-usn generated script, their in-progress USN is in the shared db; if someone else comes along and publishes a USN, pushing the shared USN db publicly, the in-progress USN information will be included as well, potentially making it show up in oval data and via landscape, even though the packages are not published.

The ideal solution IMO would get us out of updating a monolithic db as part of update preparation, and possibly it would be the publish-usn-to-website step of the publication process that that pushes things into the common db/json files.