[OVN] neutron agent not cleaned up at departure causes new units to fail

[Pedro Guimarães]

$ openstack network agent list shows:

| juju-d34d64-25-lxd-6.DOMAIN | OVN Controller agent | juju-d34d64-25-lxd-6.DOMAIN | | XXX | UP | ovn-controller |

In which the agent is marked as gone.
Indeed, that is corresponding to an octavia unit (25/lxd/6) which has been previously removed.
I can also see it on Chassis table: https://pastebin.ubuntu.com/p/pSSNqr5bBC/

As well as other ovn-chassis units that have been previously removed.

I can also see it in Encap table:
_uuid : 157c9a73-0ddf-4091-b7f2-9ea10241103c
chassis_name : juju-d34d64-23-lxd-6.DOMAIN
ip : "IP" <<<<<-----------------------------------
options : {csum="true"}
type : geneve

My issue is that, redeploying a new unit for octavia (in this case, 25/lxd/10) does not eventually work, since the ovn-chassis charm accompanying it cannot register itself.
In the /var/log/ovn/ovn-controller.log, I can see entries every minute approx. with:

2021-06-15T14:34:21.321Z|00010|ovsdb_idl|WARN|transaction error: {"details":"Transaction causes multiple rows in \"Encap\" table to have identical values (geneve and \"IP\") for index on columns \"type\" and \"ip\". First row, with UUID 0c7b68b5-3238-4a4c-883f-238590c18f37, was inserted by this transaction. Second row, with UUID 157c9a73-0ddf-4091-b7f2-9ea10241103c, existed in the database before this transaction and was not modified by the transaction.","error":"constraint violation"}

According to:
https://bugzilla.redhat.com/show_bug.cgi?id=1946179

That happens because "IP" is being reused, and given the automation (in our case, the charms) did not clean the previous entries; then this new ovn-chassis unit cannot register itself.

A comment on the bug above uggests that ovn-sbctl should be used to clean those entries.
The equivalent neutron agent should also be cleaned up.