add support for detached LUKS header and UUID fallbacks

[PapiJalopy]

Two hard failures prevented 41-snapshots-btrfs from generating a snapshot submenu when the root LUKS header is detached and cryptdevice= uses a /dev/disk/by-id path:

• grub-probe --target=fs_uuid aborted on detached headers.
• grep-based extraction of UUID from GRUB_CMDLINE_LINUX_DEFAULT failed when cryptdevice= did not contain “UUID=…”.

This patch:

1. Wraps grub-probe in a try/blkid/lsblk cascade that always returns the filesystem UUID or prints a clear error.

2. Replaces the fixed “cryptomount -u $(grep …UUID=…)” line with logic that:
   • accepts both UUID=… and /dev/disk/by-id/… syntaxes,
   • resolves paths to a canonical UUID with blkid,
   • emits ‘cryptomount -u ’ when possible,
   • falls back to ‘cryptomount -a’ only if no UUID can be extracted.

3. Keeps the previous behavior unchanged for unencrypted systems or for installations that already worked.

Result: snapshot menu is produced and boots correctly on standard (setup with inline header), detached-header, and by-id configurations; no regression for existing users.

Please read Issue #377

[Antynea]

Thank you very much for that.
Can anyone affected by this problem confirm that everything works with this patch, before merge.

[hyperreal64]

I have a weird disk setup where the Btrfs filesystem spans two disks. So the root_uuid variable only gets set when the command is as follows:

# Fallback when grub-probe fails (encrypted container, detached header...)
if [ -z "$root_uuid" ]; then
    root_uuid="$(blkid -s UUID -o value "$root_device" | tail -n1 2>/dev/null)"
fi

Otherwise, I can confirm this PR works on my system (Fedora 42).

[range4-skyz]

Could this patch be merged? I encountered the same issue when testing LUKS-encrypted Btrfs partitions on a virtual machine (archlinux-2025.12.01-x86_64.iso); this patch does indeed fix it.