On API restart, cluster nodes now re-register their primary in the local
ha_nodes and reload nftables so @peer_ipv4 is correct after a package
update or reboot without requiring a re-join.
Also fixes duplicate ha_nodes rows: preRegisterPrimary previously used
time.Now().UnixNano() as node ID, creating a fresh row each call.
Now uses a deterministic ID derived from the FQDN so repeated upserts
are idempotent.
PrimaryFQDN is now persisted in setup.json during CompleteAsNode so the
startup sync knows which primary to contact.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
de28523708