Add fastapi-ai-guardrails extension

[ulises-jeremias]

Problem

The proposal includes guardrails for input/output validation. This is a reusable AI
application concern and should not be tangled with a single chat implementation.

Solution

Create extensions/fastapi-ai-guardrails compatible with fastapi-backend.

Scope:

• Add typed guardrail hooks for input and output checks.
• Include simple defaults: max length, blocked patterns, structured response checks,
  and extension points for project-specific policies.
• Add tests for allowed and blocked inputs/outputs.
• Add docs:
  • template/docs/AI_GUARDRAILS_GUIDE.md
  • template/docs/README.md.append
• Add partial pyproject.toml only if runtime dependencies are truly needed.

Constraints:

• Do not claim this is complete AI safety coverage.
• Do not use prompt-only guardrails as the enforcement mechanism; code-level checks
  must enforce behavior.
• Do not require external moderation APIs for default tests.

Acceptance criteria:

• Generated app exposes reusable guardrail functions/classes.
• Docs explain limitations and how to extend checks.

[ulises-jeremias]

Reopened after fully reverting #93 via #94. Work remains for the assigned contributor — please continue from this issue.