Refine a run until on-brand, within budget

Bounded generate → score → refine loop. Repeats score → refine (re-rolling with the specific deviations fed back in) on a completed run and stops the instant brandFit >= minScore, iterations >= maxIterations, or the credit budget would be exceeded — whichever first. Returns the best result and exactly how many iterations and credits it took. The credit cap is never exceeded.