workflow-authoring: warn that re-running a subset of reviewer lenses keyed by a free-text schema field silently verifies nothing

A verify/fix workflow re-ran only the reviewers that had found blocking issues. It selected them by matching each reviewer's self-reported free-text `lens` field (for example `"regression-on-existing-views-plus-battery"`) against the script's own lens keys (`"regression"`). Nothing matched, so the failed lens was replaced by `null`, the blocking-findings count fell to 0, and the workflow went on as if verified. A check ran, reported success, and measured nothing.

The Quality patterns section of the workflow-authoring reference covers verify/fix loops but has no warning for this shape. Suggested addition:

> When a verify/fix loop re-runs a subset of lenses, key the subset by the script's own lens key — carry it in a closure, or use an enum-constrained schema field whose permitted values are exactly those keys — never by a free-text field the reviewer fills in. Treat an empty re-run set as an error, never as zero blocking findings.

Start in the Quality patterns section of the workflow-authoring reference, which covers verify/fix loops. Add the proposed warning about using script-owned lens keys and treating an empty re-run set as an error; the work is done when this guidance is included in that section.

Written by the indexing model from the issue text.