boot: an overlay is mounted once its NBD device has a size

[aledbf]

On kvm-runner, every report after the first boot failed with:

sudo mount /dev/nbd0 ...: can't read superblock on /dev/nbd0

A diagnostic run on the runner showed why:

[11109.963] block nbd0: NBD_DISCONNECT                         <- the previous boot's overlay
[11110.415] EXT4-fs (nbd0): unable to read superblock          <- our mount, device still 0 bytes
[11110.422] nbd0: detected capacity change from 0 to 2129920   <- the size, 7 ms later

qemu-nbd --connect returns before the kernel knows the device's size. The wait before the mount asked blkid, which answers from its cache: it reported TYPE="ext4" while blockdev --getsize64 returned 0. On a faster host the reconnect usually wins the race, which is why it never showed up locally.

Change: editOverlay now waits for a non-zero size and probes with blkid -p, which bypasses the cache.

Check: with this branch, report.yml on kvm-runner against v20260928.02 went green (run):

• 48 rows without hotplug and the 10 image variants all boot.
• The 48 hotplug=on rows are refused because that release's QEMU predates qemu: virtio-scsi and scsi-hd, and every build step fetches its own source #24 ('virtio-scsi-pci' is not a valid device model name), as expected.
• Host: Ryzen 9 5900X, kernel 6.17, echo login.

task lint passes.

🤖 Generated with Claude Code

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}