iroh 1.3, and a socket that survives a service restart

iroh 1.0.2 -> 1.3.0 (relay reconnects that keep answering, rate-limit
notices from relays, the n0 DNS resolver with configurable fallbacks,
relay connection metrics), iroh-mdns-address-lookup 0.4 -> 0.6,
n0-future 0.1 -> 0.3. No code changes were needed; the suite passes.

varde.service now keeps /run/varde across restarts
(RuntimeDirectoryPreserve=yes): stopping the service removed the
directory and with it the socket unit's file, leaving varde.socket
"active" on an unlinked path after any service restart. The install
notes say to restart the socket, which cycles both.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

dist/arch/varde.install

@@ -5,6 +5,6 @@ post_install() {
 }
 
 post_upgrade() {
-  echo ":: Restart a running daemon to pick up the new version:"
-  echo "   systemctl --user try-restart varde.service   (and/or: systemctl try-restart varde.service)"
+  echo ":: Restart a running daemon to pick up the new version - the socket, not the service:"
+  echo "   systemctl try-restart varde.socket; systemctl --user try-restart varde.socket"
 }

dist/varde.service

@@ -16,6 +16,10 @@ Restart=on-failure
 DynamicUser=yes
 StateDirectory=varde
 RuntimeDirectory=varde
+# The socket unit's file lives in this directory: removing it on every
+# service stop (the default) left varde.socket listening on an unlinked
+# path after any restart - "active" with nothing to connect to.
+RuntimeDirectoryPreserve=yes
 NoNewPrivileges=yes
 ProtectSystem=strict
 ProtectHome=yes