Move 'sleep 1' before mkfs into attach_image()

I just encountered a similar issue as #45 when creating an image with
UEFI (which implies creating a partitioning table) – sfdisk failed due
to nbd device not being ready. I think the sleep needs to be right after
`attach_image`, i.e. sooner than `mkfs`. There’s already one `sleep 1`
after writing the partitioning table which is before mkfs.

Related to 9ebdf2f

Author: jirutka

alpine-make-vm-image

@@ -202,8 +202,10 @@ attach_image() {
 	} || die 'No available nbd device found!'
 
 	qemu-nbd --connect="$disk_dev" --cache=writeback \
-		${format:+--format=$format} "$image" \
-		&& echo "$disk_dev"
+		${format:+--format=$format} "$image" || return 1
+
+	sleep 1  # see #45
+	echo "$disk_dev"
 }
 
 # Prints UUID of filesystem on the specified block device.
@@ -554,8 +556,6 @@ fi
 # useless for NBD image and prints confusing error).
 [ "$ROOTFS" = ext4 ] && mkfs_args='-O ^64bit -E nodiscard' || mkfs_args='-K'
 
-sleep 1  # see #45
-
 mkfs.$ROOTFS -L root $mkfs_args "$root_dev"
 
 root_uuid=$(blk_uuid "$root_dev")