fix: keep the eth0 direct-cable static up when DHCP has no server

[ArnavVWadhwa]

On a direct laptop-to-Pi cable, eth0 was tearing down and re-activating every 45s, dropping the 172.17.21.2 static each cycle. Pinging the Pi times out for several seconds, then recovers, on repeat.

Cause: the direct link has no DHCP server and no router, so ipv4=auto (DHCP) and ipv6=auto (RA) both produce no config. NetworkManager then fails the whole activation as ip-config-unavailable and autoconnect restarts it, so may-fail=true on ipv4 never gets a chance to help. Confirmed on the Pi (Trixie, NM 1.52):

dhcp4 (eth0): beginning transaction (timeout in 45 seconds)
state change: ip-config -> failed (reason 'ip-config-unavailable')
Activation: failed for connection 'dpea-eth0'
state change: failed -> disconnected
policy: auto-activating connection 'dpea-eth0'   # loops every ~45s

Fix: set ipv6.method=link-local so IPv6 always activates successfully. That keeps the connection up, the ipv4 DHCP attempt fails quietly under may-fail=true, and the static stays pinned. DHCP still wins when the Pi is plugged into a real network.

Address unchanged (172.17.21.2/22). No workflow or CI changes.