Add CodSpeed performance benchmarks

[codspeed-hq]

What this does

This PR integrates CodSpeed for continuous performance tracking. It adds benchmarks for performance-critical utility functions and a GitHub Actions workflow that runs them on every push to main and on pull requests.

Changes

Benchmark harness

• Installed @codspeed/vitest-plugin in the src workspace
• Created src/vitest.bench.config.mts with the CodSpeed vitest plugin configured in simulation mode
• Added 18 benchmarks across 3 files covering core utility functions:
  • Text normalization (src/benchmarks/text-normalization.bench.ts): normalizeString and unescapeHtmlEntities with varying input sizes and option combinations
  • Tool ID sanitization (src/benchmarks/tool-id.bench.ts): sanitizeToolUseId, truncateOpenAiCallId, and sanitizeOpenAiCallId covering clean IDs, special characters, and long inputs requiring hash truncation
  • Command parsing (src/benchmarks/parse-command.bench.ts): parseCommand with simple commands, chained operators, pipes/redirections/subshells, multi-line scripts, and shell variable expansions

CI workflow

• Added .github/workflows/codspeed.yml that runs on push to main, pull requests, and manual dispatch
• Uses the existing setup-node-pnpm composite action for consistent dependency installation
• Runs benchmarks in CodSpeed simulation mode (deterministic CPU simulation, less than 1% variance)
• Uses OIDC authentication (no secrets required)

README

• Added CodSpeed badge alongside existing project badges

How to run benchmarks locally

cd src && npx vitest bench --config vitest.bench.config.mts --run

Next steps

• Review the initial benchmark results on the CodSpeed dashboard
• Add benchmarks for additional hot paths as needed (e.g., message consolidation in @roo-code/core, JSON schema validation)
• Consider enabling performance checks to flag regressions on pull requests

[codspeed-hq]

Congrats! CodSpeed is installed 🎉

🆕 18 new benchmarks were detected.

You will start to see performance impacts in the reports once the benchmarks are run from your default branch.

Detected benchmarks

• all options enabled: 850.8 µs
• ID with special characters: 23 µs
• simple command: 142.9 µs
• complex command (pipes, redirections, subshells): 264 µs
• multi-line script: 352.7 µs
• short input: 42.6 µs
• shell variables and expansions: 218.4 µs
• short ID (no truncation): 18.5 µs
• clean ID (no-op): 32.3 µs
• chained commands (&&): 155.1 µs
• short clean ID: 18.6 µs
• long ID (needs truncation + hash): 117.9 µs
• short string: 66.3 µs
• long ID with special characters: 130 µs
• special characters ID: 24.2 µs
• long string (100 lines): 842.6 µs
• smart quotes only: 252.6 µs
• long input (50 repetitions): 231.4 µs

---