feat: hard persona gate enforcement + interactive CLI (v1.7.0)

- Replace readline with @inquirer/prompts (arrow-key select, checkbox multi-select)
- Persona switching now requires 'persona exit' first — no bypass allowed
- Persona gate fails with exit(1) instead of exit(0)
- Gate double-checks prerequisite artifact files exist on disk (not just config flags)
- All agent SKILL.md prerequisite checks changed to hard STOP with no proceed-with-assumptions option
- Node engine requirement bumped to >=18 (Node 16 EOL)

chore(release): bump version to 1.6.0

- Add QA persona (agent-qa, test-plan, test-execution, bug-triage, qa-signoff)
- Add Release persona (agent-release, release-checklist, deploy-runbook, release-notes)
- Add Team Lead agent skill (agent-lead) with sprint-status and retrospective
- Add review-adversarial core skill (available to all personas)
- Add reenter command for backward phase re-entry with audit logging
- Add audit trail (lib/utils/audit.js, _superml/audit.log)
- Add persona status and persona exit sub-commands
- Add market-research and technical-research skills (Phase 1)
- Add check-readiness skill (Phase 3)
- Add Azure DevOps, GitLab, and Confluence sync integrations
- Sync modernize_artifacts path in init.js config template
- Update all docs: personas, workflows, skills-reference, README
- Add IDEA-BEHIND.md with full system design rationale

Release v1.5.3

Release v1.5.2

ci: add npm publish workflow + GitHub Pages UI