[Geo] Prevent using non-empty tracking database

It seems like there are some flows when it sounds reasonable (at least, at first glance) to use an already existing tracking database. In this case, the secondary node will be invalid and sometimes it's hard to debug. I propose to add a special check into our geo health checker. The first thing that comes to my mind is to check the oldest record in the tracking database and compare it with GeoNode record's created_at.