Ignore IRRd v2 / v3 clutter in dump files

[job]

Describe the bug
IRRd v4 sometimes has trouble loading dumps from older IRRd versions. Older IRRd instances would mark deleted objects with *xx as the first three chars of the object type, which meant that later on down in the file there will be a newer version of that object. This way IRRd wouldn't have to reshuffle its full flat text file database all the time.

To Reproduce
Today's RADB dump can't be loaded:

Error occurred while processing object:
Unknown object class: *xxte

Expected behaviour
The irrd_load_database command should be able to ingest dump files that have IRR objects of object class types such as *xx-num:,*xxner:, *xxset:, *xxte-set:, *xxte6:, or *xxte:), and just ignore those without warning and not error. Just pretend that object didn't exist.

IRRd version you are running
E.g. 4.0.1

Additional context
Add any other context about the problem here.

[mxsasha]

Yes, we can do this, but see my comment #233 as well. As in, I'm not sure why you're loading RADB manually instead of through the regular mirror process.

[job]

Mirroring had broken for some reason, and the NRTM gap was too large for the source IRRd instance to fill so it wouldn't re-established, so we had to re-seed from RADB's FTP dump which contains these *xx prefix'ed object classes.

I agree with the tending to be strict, but in the case of *xx objects we know what they are and why they are there, we can safely ignore those objects.

[mxsasha]

As https://irrd4.readthedocs.io/en/latest/users/mirroring.html now says:

Unknown object classes that start with *xx are silently ignored, as these are harmless artifacts from certain legacy IRRd versions.

[job]

this works as advertised!