feat: add case tag rollups to gate reports

[brandonwise]

Summary

• add optional tags on cases so specs can group related checks by workflow or surface area
• roll tag summaries into JSON + Markdown reports with pass rate, passed/total, and failing case ids
• cover the new behavior in evaluator tests and show tags in the example spec + README

Why now

• Reddit demand keeps landing on actionable eval operations rather than raw scores: How do you automate LLM evals? and how are people actually trusting LLM eval scores in production?
• X keeps pushing eval-driven development as the current build discipline, not a side topic: Vercel on eval-driven development and AI eval flywheel notes from ai.engineer
• GitHub trending is crowded with agent infra and code-quality tooling (GitHub trending), while Product Hunt's AI-content-detection/humanizer lane is already saturated (category); this makes a sharper eval-actionability increment a better positioning move today

Validation

• python3 -m pytest tests/test_evaluator.py -q — PASS
• python3 -m pytest -q — PASS
• python3 -m ruff check src tests — PASS
• python3 -m build — PASS
• PYTHONPATH=src python3 -m agent_release_gate.cli evaluate --spec examples/spec.yaml --results examples/results.json --markdown /tmp/argate-smoke-VAau0J/report.md — PASS, Markdown report includes the new ## Tag summary section
• node /Users/bwise/.openclaw/workspace/projects/humanizer/src/cli.js score --file README.md --ignore-code — PASS (🟢 1/100)

Risk / rollback

• low risk: tags is optional and existing specs still parse unchanged
• rollback is a single revert of this PR if downstream consumers dislike the extra summary fields