[Refactor] Collapse sleep-check cancel-vs-fail finalization

[roomote-roomote]

Created by Roomote. Follow up by mentioning @roomote or in the web UI.

What changed

Sleep-check recovery paths that finalize a swept job as canceled or failed no longer each inline the same status resolution, event recording, logging, and finishCloudJob sequence. Those branches now call a shared finalizeSweptJob helper and only supply path-specific messages and details.

Why this change was made

After the canceled-task recovery work in #53, timed and heartbeat recovery paths still duplicated nearly identical cancel-vs-fail finalization blocks in an already large scheduled job. The next recovery path would have copied that third branch again.

Impact

No intentional user-facing behavior change. Canceled vs failed status selection, sleep-check events, and finish arguments stay the same; the scheduled job is easier to extend without another copy-paste when a new sweep recovery path is added.

[roomote-roomote]

1 issue outstanding. See task

• Medium: apps/bullmq/src/scheduled-jobs/sleep-check.ts:677-689 preserves a terminal sleep-check event and emits the success-style warning before finishCloudJob succeeds. For both heartbeat recovery branches this reverses the previous order, so a database/read or transaction failure in finishCloudJob can leave the run active while the durable audit trail already says it was failed or canceled; keep those outcome signals after successful finalization (or explicitly handle the failure case) before treating this as behavior-preserving.