Iteration Budget Math

ISO C. Implement size_t iters_for_budget(double budget_ms, double per_iter_ms) — return the number of whole iterations that fit the budget, rounding up (a partially affordable iteration still runs once); return 0 when either argument is <= 0. Implement double bench_sink(size_t iters) that accumulates a workload into a local volatile double sink (sink += (double)i / 4.0 each iteration) and returns the sink — proving the work was not optimized away.