iikit-clarify needs eval scenario coverage

[jbaruch]

The iikit-clarify skill has zero eval scenarios despite carrying substantial decisional logic:

• Auto-detection of target artifact across 6 artifact types (spec, plan, checklist, testify, tasks, constitution)
• Ambiguity scanning per artifact-specific taxonomy
• Sequential questioning + answer integration
• Per-artifact reference format (FR-xxx, US-x, SC-xxx, plan headers, check items, T-xxx, principle names)
• Testify exception path (Gherkin can't hold markdown)
• Checklist re-validation after spec clarification (added in feat(iikit-clarify): re-validate checklists after spec clarification #79)

Per jbaruch/coding-policy: plugin-evals: "Every skill with decisional logic ships eval cases."

Surfaced on PR #79 when the policy reviewer flagged the new branching behavior — but the underlying issue is that the whole skill has been untested in evals. Filing this for a focused eval-coverage effort rather than tacking one scenario onto every behavioral change PR.

Scope

Add scenarios covering at least:

• Auto-detection picks correct artifact (positive case per artifact type)
• Question generation respects per-artifact reference scheme
• Integration writes Q&A back into the right section format
• Testify exception writes to tests/clarifications.md not the .feature file
• Checklist re-validation (from feat(iikit-clarify): re-validate checklists after spec clarification #79): stale-regression, resolved-gap, non-spec no-op

Use tessl scenario generate for happy-path then hand-author negative cases per plugin-evals guidance.