Make preview totals describe the rows actually emitted

In a UI acceptance preview, do not place full-population header totals above a sampled row fixture unless the sampling boundary is unmistakably labeled. Prefer deriving tab totals from the exact dataset emitted to the preview, and test that all displayed queue partitions sum to the visible active count. If a preview intentionally combines population metrics with sampled rows, label both scopes separately so reviewers cannot mistake the fixture for the population.