docs: add AI workflow safety overview

[ProfRandom92]

Adds a concise README overview of the local AI workflow safety/evidence chain.

Includes:

• safe_pr_gate
• agent_artifact_bundle
• validate_agent_artifact_bundle
• pr_body_from_agent_bundle
• ai_workflow_snapshot
• MCP context output references

Scope:
Documentation only. Conservative positioning: local-only, deterministic, artifact-backed, no external APIs, and not an autonomous agent framework.

Validation:

• No tests run; README-only change.

[gemini-code-assist]

Code Review

This pull request updates the README.md file to include a new section titled "AI workflow safety evidence." This section describes a local-only, deterministic evidence chain for AI-assisted repository work, highlighting several scripts used for safety gating, artifact bundling, and validation. I have no feedback to provide as there are no review comments.