Route order preference

[mxsasha]

Adds a new route order preference option, which can suppress route objects if an overlapping higher priority route object exists. Also restructures the docs a bit around this feature.

Docs: https://irrd.readthedocs.io/en/ropref/admins/object-suppression/ and https://irrd.readthedocs.io/en/ropref/admins/route-object-preference/

Some performance indication

Time taken is proportional to the number of objects that change state.

Numbers from my (Celeron J4115, 32GB, scavenged SATA SSD) testing box:

• Moving from disabled to ALTDB at 50, RADB at 100, others at 900 (typically only done once):
  • 1.099.850 moved to suppressed
  • 7 minutes to scan objects, decide new states, enrich object to be changed
  • 3 hours to generate journal entries for newly suppressed objects
  • 1h28m to update the state of the now suppressed objects
  • 12s to commit the transaction
• Then, changing ALTDB to 150:
  • 9.800 objects moved to visible, 20.000 objects moved to suppressed
  • 2m15s for scan, deciding on new state, enriching, and adding journal entries
  • 33s for updating state
  • 2s to commit the transaction
• Periodic full update when there are no state changes: about 2 minutes for scanning and deciding new state.
• Partial update after e.g. processing an NRTM update: several seconds or shorter.

However, it's questionable whether it makes sense to even enable the journal on these large state changes. NRTMv3 will not scale to catching up with 1 million objects. Therefore, while it may be possible to speed up the journal insert time, it is not particularly useful.

Todo

• Enable by default on query side
• Support single validator queries
• Fix conflict with Increase journal insertion speed #731 (minor)
• Schedule periodically
• Run validator after database changes
• Make sure this behaves correctly when disabled
• Ensure multi-preference overlaps work as intended
• Overlaps with RPKI invalid routes?
• Code TODOs
• Docstrings
• Document log messagee
• Is the trigger specific to route objects or perhaps also picking up inet6num?
• Don't trigger a status check for objects where we just changed the route preference
• Is this in database status call? !J? Status page?
• Is query doc up to date with new !f ?
• Import timer?
• Fix up the pyradix dependency
• Do some real data testing
• Validate new setting
• Documentation
• Review naming
• Refactor/review
• Remove obsolete method