Make evals fast and hang-proof (timeout, retry, concurrency) (#44)

The "Run evals" step ran 24 API calls sequentially with no request timeout, so
it was slow and could stall indefinitely if one call hung.

- bin/lib/anthropic.mjs: per-request timeout (120s) via AbortController + retry
  (2x, backoff) on 429/5xx/timeout. Fails fast on 4xx (bad key/model).
- evals/run-evals.mjs: run (case × model) tasks through a concurrency pool
  (default 4, --concurrency to tune); preserves result order.
- eval-leaderboard.yml: job timeout-minutes: 20 as a safety net.

Applies to the next run. The hardening also benefits the Action runner and
`generate`, which share the client.


Claude-Session: https://claude.ai/code/session_016JWn5jRD5tcEFKrubjQ6Px

Co-authored-by: Claude <noreply@anthropic.com>

CHANGELOG.md

@@ -9,6 +9,12 @@ each new wave of skills bumps the **major** version, extensions and fixes bump
 
 ## [Unreleased]
 
+### Changed
+- **Faster, hang-proof evals.** The Anthropic client now has a per-request timeout (120s)
+  and limited retries (429/5xx/timeout); the eval harness runs cases concurrently
+  (default 4). The leaderboard workflow has a 20-minute job timeout. A 24-call run that
+  was sequential now finishes in a few minutes and can't stall a job indefinitely.
+
 ### Added
 - **One-click leaderboard updates in CI** — `.github/workflows/eval-leaderboard.yml`
   ("Update Skill Leaderboard") runs the evals with the `ANTHROPIC_API_KEY` secret, commits