Add AI Agent Security section with TokenFence

[u4ma-kev]

As AI agents become more prevalent in Python applications, runtime security for LLM-powered agents is a growing concern. This PR adds a new 'AI Agent Security' section with TokenFence — an open-source runtime guardrails SDK.

What TokenFence provides:

• Per-workflow budget caps to prevent cost overruns
• Automatic model downgrade when approaching limits
• Kill switches for runaway agent spending
• Least-privilege policy engine (allow/deny/require_approval) with audit trail

Zero dependencies. MIT licensed. Available on PyPI: pip install tokenfence

• GitHub: https://github.com/u4ma-kev/tokenfence-python
• PyPI: https://pypi.org/project/tokenfence/
• 104 tests passing