feat(multishot): expose output budget controls

[drewstone]

Summary

• add multishot agent/tool-followup/driver/judge max-token controls
• add simulated-user driver fallback model support
• bump agent-eval version trio to 0.96.5

Verification

• pnpm test: 247 files, 2533 passed, 2 skipped
• pnpm typecheck
• pnpm lint: exit 0; reports existing warnings outside this diff
• NODE_OPTIONS=--max-old-space-size=8192 pnpm build
• git diff --check
• git merge-tree --write-tree origin/main HEAD -> 5cf4700e23edcfe8d477c140085617a4b56c5ff2

[tangletools]

✅ Auto-approved PR — c7f03fa6

Blanket team auto-approval is enabled for this reviewer service.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: blanket_auto_approve · 2026-06-22T22:22:01Z}