Settle.elapse constructor

Spend all of budget on the fake clock, whatever the frame loop is doing — the policy for work that waits on a timer rather than on frames.

Where Settle.upTo asks the app whether to keep going, this one does not ask: it advances the clock, so a timer due inside budget fires and whatever it wakes gets its frames. Pumping past a quiet screen is close to free — a pump with nothing due builds nothing — so the cost is not the frames.

The cost is that the clock really does move. A snackbar that auto-dismisses at four seconds is gone by the end of a five-second budget, and a screen captured after this policy is the screen as it stands at budget, not as it stood when it stopped changing. That is why it is not the default: reach for it where the question is does this work at all rather than what does it look like now.