feat(CI-CD): shard mutation-dashboard.yml so its runtime stops scaling with the helper count

[baxyz]

Summary

• Replaces mutation-dashboard.yml's single full-project mutation job with a 4-way matrix (mutation-shard), each shard testing a deterministic, self-rebalancing round-robin subset of helper files (sorted file list, (NR-1) % SHARD_COUNT == shard) — no category list to maintain as helpers are added or removed.
• New aggregate job downloads all shard reports, merges them (scripts/mutation/merge-shards.ts), uploads to the Stryker Dashboard, and commits the refreshed reports/mutation/incremental.json baseline — same responsibilities the old single job had.
• Each shard's own report already covers the whole project (Stryker's incremental mode carries forward every file outside a shard's --mutate scope from the committed baseline) — verified by reading Stryker's incremental-differ.ts and json-reporter.ts source directly. The merge script uses each shard's manifest to pick the one shard that actually tested each file fresh, rather than a stale carried-forward duplicate.
• Hard-fails loudly if a shard's own report is missing one of its assigned files, or if fewer shard artifacts exist than SHARD_COUNT (a shard job failing before its upload step would otherwise let download-artifact's glob pattern silently return an incomplete set) — both paths verified locally with synthetic fixtures.
• Fixes a real bug caught during testing: writing the shard's file list with echo left a trailing newline baked into the last filename once re-read via fs.readFileSync in a later step; switched to printf.
• Closes Shard mutation-dashboard.yml so full runs stay under a hard timeout #159.

Why

mutation-dashboard.yml's runtime grew from 22 to 40+ minutes as helpers were added today, needing a manually bumped timeout (#160) — a fixed cap that would only be outgrown again as the project keeps growing. Sharding keeps each job's own runtime roughly constant regardless of total helper count.

Test plan

• YAML validated with js-yaml
• pnpm run typecheck / pnpm run lint clean
• merge-shards.ts verified locally against synthetic shard fixtures: correct fresh-vs-stale file selection, missing-file-in-report failure, and missing-shard-artifact failure
• Next push-to-main run completes with all 4 shards + aggregate succeeding, and the incremental baseline gets committed
• Confirm the merged baseline still functions correctly as an incremental-mode baseline on the run after that (i.e. mostly-skipped, fast)

🤖 Generated with Claude Code

[github-actions]

✅ PR Validation Passed

All checks passed!

---

📋 Pipeline Status

	Job	Status
✅	🔢 Version	passing
✅	🏗️ Build	passing
✅	🧪 Tests	passing
✅	📝 Lint	passing
✅	📘 TypeCheck	passing
✅	🔐 Security Audit	passing
✅	🧾 Conventional Commits	passing
✅	🔗 Coherency	passing

---

📊 Code Coverage

✅ Overall Coverage: 100.0% — Target reached! 🎯

	Metric	Progress	Coverage
✅	Lines	▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓	100%
✅	Branches	▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓	100%
✅	Functions	▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓	100%
✅	Statements	▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓	100%

---

🧬 Mutation Testing

⏭️ Mutation Testing: skipped

---

🌐 Runtime Compatibility

✅ Runtime Compatibility: 3/3 — All runtimes passed!

	Runtime	Status
🟩	Node.js 26 (LTS) v26.8.1	✅
🦕	Deno 2.8.3	✅
🍞	Bun 1.4.2	✅

🔄 Runtime compatibility is informational only and does not block the PR

---

⏱️ Benchmarks

⏭️ Benchmarks: skipped

---

ℹ️ About this report

• 🎯 Coverage Target: 100% for all metrics
• 🧬 Mutation Testing: powered by Stryker (non-blocking)
• ⏱️ Benchmarks: powered by Vitest Bench (non-blocking)
• 🔄 This comment updates automatically with each push
• 📈 Coverage is measured using Vitest + v8

---

_{🤖 Generated by @helpers4 CI • 2026-09-08}