feat(personas): Add agent personas for specialized prompt injection (v3.3)

Add agent personas that inject domain expertise and behavioral guidelines
into prompts before agent execution. This improves response quality by
giving each agent role-specific instructions.

New personas:
- backend-architect: API design, microservices, database patterns
- security-auditor: OWASP, vulnerability scanning, compliance
- frontend-architect: React/Next.js, component design, accessibility
- researcher: Deep investigation, pattern analysis, synthesis
- reviewer: Code quality, security review, best practices
- implementer: Clean code, TDD, production-ready implementation
- synthesizer: Result aggregation, conflict resolution

Integration:
- Double Diamond phases auto-select appropriate personas
- spawn_agent/run_agent_sync accept optional role/phase params
- Configurable via --no-personas flag or DISABLE_PERSONAS env var

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

Author: nyldn