## Problem

`omarchy-hibernation-setup` computes `resume_offset` correctly at setup time (#13768 further fixes the stale-file path on setup/remove). But between setups, the swapfile's physical location can move (balance, defragmentation) and nothing detects it:

- Hibernate *writes* the image correctly — swsusp goes through the swap extent map.
- On resume, the initramfs reads the image header from the literal `resume_offset` physical offset; btrfs is not mounted yet, so the extent map does not apply. A stale offset means the `S1SUSP` signature check fails, the kernel gives up, and the machine cold-boots: session lost, nothing surfaced anywhere except the absence of a resume.

Hit exactly this on a current Omarchy install: configured `resume_offset=1630758`, actual swapfile location `1894950`.

## Proposal

A tiny oneshot installed by `omarchy-hibernation-setup`, comparing the running kernel's offset against the live swapfile at every boot and warning loudly on mismatch:

```bash
EXPECTED=$(cat /sys/power/resume_offset)
ACTUAL=$(btrfs inspect-internal map-swapfile -r /swap/swapfile)
[[ $EXPECTED == "$ACTUAL" ]] && exit 0
logger -p err -t resume-offset-check "resume_offset stale: kernel=$EXPECTED swapfile=$ACTUAL"
# + notify-send to the desktop session
exit 1
```

Cost is one `map-swapfile` call per boot. Complementary to #13768: that PR makes setup/remove correct, this makes drift visible. Happy to send a PR in repo style (test/shell.d + AGENTS.md conventions) if there's appetite.