fix(status): align stall alerts with index schedule

[randomblocker]

What changed

• derive the default cursor-stall threshold from the production reindex schedule
• allow two valid index intervals before declaring a cursor stalled, with a five-minute floor and 24-hour ceiling
• preserve an explicit valid DACS_CURSOR_STALL_SECONDS override
• fall back to the schedule-aware default when either configuration is invalid
• document the index interval and stall override together

Why

Production defaults to a 15-minute indexing interval but status used a fixed five-minute stall threshold. The live endpoint therefore reported cursorStalled: true about six minutes after a successful pass, before the next scheduled pass was due. That made healthy periodic operation look unhealthy for much of every cycle.

With the default schedule, the threshold is now 1,800 seconds. Operators can still choose an explicit tighter threshold when their deployment runs more frequently.

Validation

• npm run typecheck
• all eight status/dead-letter diagnostic subtests emitted passing results locally
• focused schedule assertions cover the default interval, short/custom intervals, explicit override, and invalid override fallback
• git diff --check

Closes #75