fix(jobs): don't busy-loop ownstatus on a persistent error (audit M12)

[passcod]

Fixes M12 (medium) from the audit in #370.

The bug

The loop slept at the bottom and continued on error, so the error path skipped the sleep entirely:

loop {
    if let Err(err) = record_own_status(…).await {
        error!(…);
        continue;          // ← straight back to the top
    }
    sleep(Duration::from_secs(60)).await;
}

A persistent, fast-failing error — the database unreachable, a missing Server::own row — then spins as fast as the query can fail: thousands of connection attempts and error! lines a minute, hammering the database it's already failing to reach and burying the logs.

The fix

The sleep moves to the top, which is where the sibling pingtask puts it for exactly this reason. There's nothing left to continue past, so the error path just logs and falls through to the next tick.

One behaviour change worth noting: the first status is now recorded 60s after start rather than immediately. That matches pingtask, and the job's own cadence is a minute, so nothing downstream distinguishes the two.

No test: this is a loop { sleep } in a spawn, and asserting on it needs time control the audit rates as moderate-to-hard for a one-line ordering fix. The change is a straight comparison with the sibling job.

---

Generated by Claude Code