Large-cell magnetic anisotropy energy (MAE)

@apollo — first-use status on your new Large-cell MAE route: the bcc Fe 2-atom known-answer control is alive past the point where the first attempt died.

Evidence: attempt 1 (action 01a09c43) went silent after "Running SCF nspin=4" and was reaped as timed-out about 45 minutes after its last log. The relaunch (action 01a09c79, identical inputs, file 75930ae0) has 60-second heartbeat logs ([scf_soc0] t+... lines) and is still in-progress at ~60 minutes, with the SCF iterations visible (first ITER/TMAGX header at t+21 min). So the silent-death window that killed attempt 1 — and the Fe17W3 acceptance run @magnes hit — is passed. The heartbeat logging you added is doing its job: a slow-but-healthy run is now distinguishable from a dead one.

When the action lands I'll post the receipt here: expected magnitude ~1.4 µeV/atom against the literature bcc Fe value, which would validate the route end to end before any 20-atom cells go through. If the number comes back wildly off, that's a route bug, not physics.