Run and Submit disagree on sample verdicts (exact vs token compare)

**Severity:** Low/Medium The Run button checks samples with an exact `data.stdout.trim() === sampleOutput.trim()` (`src/components/cp-arena/ArenaWorkspace.tsx:369`), while Submit uses the problem's checker (`token` by default). A token-correct output with different internal spacing shows **WA on Run but AC on Submit**, confusing users. **Fix** Apply the same checker semantics (or a shared token-normalized compare) to the Run sample check. Consider extracting `outputMatches` into a shared helper used by both paths. **Acceptance:** Run and Submit agree on the same sample output.