M27.2: Prompt templates module with persona/philosophy/guardrails blocks

[bug-ops]

Parent: #590
Blocked by: M27.1

Scope

Create zeph-core/src/prompt_templates.rs with modular prompt text blocks:

1. Persona blocks (3 variants, one per InteractionMode):

   • PERSONA_ASSISTANT, PERSONA_CODING, PERSONA_AUTOMATION

2. Philosophy blocks (3 variants, one per ExpertiseLevel):

   • PHILOSOPHY_BEGINNER, PHILOSOPHY_INTERMEDIATE, PHILOSOPHY_EXPERT

3. Guardrails blocks (up to 9 combinations, grouped by dominant factor):

   • Common security guardrails (always present)
   • Level-specific guardrails (beginner gets confirmation requirements, expert gets minimal)
   • Mode-specific guardrails (coding gets test-before-commit, automation gets error-only reporting)

4. Selector functions:

   • fn persona(mode: InteractionMode) -> &'static str
   • fn philosophy(level: ExpertiseLevel) -> &'static str
   • fn guardrails(level: ExpertiseLevel, mode: InteractionMode) -> String

Acceptance Criteria

• All text blocks as const &str (no external files)
• Prompt text in English per project conventions
• Selector functions are #[must_use]
• Unit tests verifying each combination produces non-empty, distinct output
• No duplicate content between blocks (DRY)