feat: add action summary outputs

[brandonwise]

Summary

• add scan --summary-output <path> so CI can capture score/count metadata without dumping full finding bodies
• expose GitHub Action outputs for score, grade, severity counts, scan counts, and report metadata
• publish an action job summary and extend action/integration coverage for the new path

Rationale

• Product Hunt is still clustering around AI agents + AI infrastructure, so CI-ready trust tooling is more defensible than another generic agent wrapper: https://www.producthunt.com/topics/artificial-intelligence
• GitHub demand around MCP security keeps rising, with active projects like https://github.com/google/mcp-security, https://github.com/slowmist/MCP-Security-Checklist, and https://github.com/trailofbits/mcp-context-protector
• The local TikTok/X trend pass for 2026-05-27 pointed at practical, workflow-embedded AI tooling over polished "AI slop", which makes actionable CI output a better signal than surface-level copy changes

Validation

• PASS: cargo test --all-targets
• PASS: cargo test summary_output -- --nocapture
• PASS: cargo clippy --all-targets -- -D warnings
• PASS: cargo fmt --all -- --check
• PASS: cargo build --release
• PASS: bash tests/action_smoke.sh
• PASS: node /Users/bwise/.openclaw/workspace/projects/humanizer/src/cli.js score < README.md -> 35/100