saga + saga_node_event tables are never garbage collected, but they should be

[smklein]

In the spirit of #6615 , I looked at SHOW TABLES on dogfood to see where we have high-volume tables that are missing garbage collection.

In addition to the issue reported by #6615, I noticed the following tables have unusually high row counts:

  public      | saga                                              | table    | root  |             2555404 | NULL                                                                                                                              
  public      | saga_node_event                                   | table    | root  |            55688221 | NULL

I think this is "expected, but also bad". We should have a mechanism to garbage collect the tens of millions of saga events that have been created, rather than holding onto them and wasting disk space in perpetuity.

[davepacheco]

Yeah, we definitely should clean these up.

It would be really nice if we slurped these out of the database and saved them to a flat file (which is by comparison much cheaper to store) and potentially kept those indefinitely, or at least much longer. These could be really useful for support.