install: Use sfdisk, not lsblk

[cgwalters]

This works around an issue where lsblk ends up parsing data cached from udev in /run/udev, but we don't have that mounted by default.

Adding a new mount point hard requirement is logistically complicated right now - we will eventually switch
to dynamic mounts with open_tree (cc
#380 )
but this is a relatively straightforward workaround.

[cgwalters]

 > bootupctl backend install --write-uuid --device /dev/loop0 /run/bootc/mounts/rootfs
cp: error writing '/run/bootc/mounts/rootfs/boot/efi/EFI/centos/shimx64-centos.efi': No space left on device
cp: error writing '/run/bootc/mounts/rootfs/boot/efi/EFI/centos/shimx64.efi': No space left on device
cp: error writing '/run/bootc/mounts/rootfs/boot/efi/EFI/centos/mmx64.efi': No space left on device
cp: error writing '/run/bootc/mounts/rootfs/boot/efi/EFI/centos/grubx64.efi': No space left on device
cp: error writing '/run/bootc/mounts/rootfs/boot/efi/EFI/BOOT/BOOTX64.EFI': No space left on device
cp: error writing '/run/bootc/mounts/rootfs/boot/efi/EFI/BOOT/fbx64.efi': No space left on device
error: boot data installation failed: installing component EFI: Failed to copy
ERROR Installing to disk: Installing bootloader: Task Running bootupctl to install bootloader failed: ExitStatus(unix_wait_status(256))

Hmm that's a new one. It's a bit baffling, looking at the space on the host before:

+ df -h /
Filesystem      Size  Used Avail Use% Mounted on
/dev/root        72G   34G   38G  48% /

And the rootfs size is fine:

 Creating root filesystem (xfs) on device /dev/loop0p3 (size=9.5 GB)

And the container image size is fine:

 layers already present: 0; layers needed: 74 (1.7 GB)

[jlebon]

Looks sane to me.

(util-linux/util-linux#3099 seems related here; --> util-linux/util-linux#2047 --> coreos/fedora-coreos-config#2181 (comment)).