Check non-inclusive DHCP overrides

[pajod]

networkd renamed section DHCP=>DHCPv4. Two options had been documented at the time, they include DHCPv6: https://github.com/systemd/systemd/commit/4f7331a85eee2134cc2383a9eeb0445593a6be15 - Not so most of the others. They need to move to DHCPv6 section.

Any guidelines on such breaking change? This will cause trouble with incompletely applied (yet not generating warnings) configurations that, by chance, work now, e.g.:

UseMTU=yes may causes interface reset for some hardware (e.g. Intel NIC) and infinite loop -- Originally posted by @yuwata in https://github.com/systemd/systemd/issues/23484#issuecomment-1175899449

It won't suffice to just teach combine_dhcp_overrides() that most options must and all options should go into separate sections. Admins must somehow be able to discover why their effective configuration changed.

References: man 5 systemd.network https://github.com/canonical/netplan/tree/main/src/networkd.c https://github.com/systemd/systemd/tree/main/src/network/networkd-network-gperf.gperf DHCP.UseHostname => DHCPv4.UseHostname https://github.com/systemd/systemd/commit/1346b1f0388f4100bb3c2a2bb23bc881769c020c https://github.com/systemd/systemd/pull/13006 DHCPv6.UseFQDN => DHCPv6.UseHostname https://github.com/systemd/systemd/pull/18543 https://github.com/systemd/systemd/pull/18668

[pajod]

@slyon (edit: incorrect ping) I had given up on a proper fix because systemd changed that code a few times and there was no way one netplan version could serve all of them in a simple fashion. Maybe this needs to be fixed in a only-to-be-used-with-systemd-v250+ version and after requesting systemd to document as stable everything netplan uses. Created a somewhat more bug-report-like structured report here still: https://bugs.launchpad.net/netplan/+bug/2078954