refactor(persona-quiz): regenerate DAG with stricter per-phase axis diversity

Same 156-node 15-question structure as before, but the bucket subagents
re-authored with stricter prompts:

- Declarative-only enforcement (zero interrogatives now — `Do you / Have
  you / Is your` openings are gone, no question marks anywhere)
- Per-phase axis menu (MEDIUM / AUDIENCE / SCALE / OWNERSHIP / TOOLING /
  WORKFLOW / PHILOSOPHY / CRAFT / DEPTH / PROCESS) with an explicit rule
  that no two prompts inside the same phase may probe the same axis
- Explicit guard against c1/c2 conceptual overlap in the infra bucket
  (Cloud Platform Engineer stays platform-building, SRE stays reliability)
  and clean c1/c2 + c3/c4 split in the specialty bucket (game vs embedded,
  leader vs founder)
- Self-audit step where each agent rewrites siblings that probe the same
  axis and rewrites any interrogative phrasing as a declarative statement

Verifier still passes: 156 nodes, 16 archetypes reachable, 0 invalid tags,
path length = 15 from every Q1 entry. Tests 4/4. Typecheck + lint clean.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: davidercruz