Doctor's sharing check blames the port floor when the daemon is simply stopped #42

On a machine that shares (#41), doctor's Sharing check gives the wrong advice when the daemon isn't running at all.

What happens

After a reboot nothing starts grove, so the daemon is down and nothing answers DNS on the tailnet address. Doctor reports both findings, and the Sharing one sends you looking in the wrong place:

Grove           not running
Sharing         nothing answers DNS for *.capella.grov.site on 100.76.35.70:53

Start it with grove start, which puts it in the background.

The daemon answers there while it shares. If it is running an older build, run grove restart. If port 53 is refused, run grove share again, which offers to lower the port floor. Its log, daemon.log in the state directory, says which.

The port floor and the build are both fine. The only problem is the one the Grove line already names.

What it should do

When the daemon isn't answering, the Sharing finding should defer to the Grove line: say sharing is paused until grove starts, with no advice of its own, or stay quiet. checkSharing in cmd/grove/doctor_share.go can take whether the daemon answered, as checkPort443 already does.