fix(runtime): pass backend generation options

[drewstone]

Summary

• Add optional temperature and maxTokens passthrough to createOpenAICompatibleBackend.
• Forward those options through resolveAgentBackend for router/tcloud/cli-bridge paths.
• Map maxTokens to OpenAI-compatible max_tokens and preserve the existing omitted-field behavior when unset.
• Regenerate API docs for the public option surface.

Why

This closes #460 and lets product/lab callers use the shared backend resolver without losing generation settings needed for fair direct-chat controls.

Verification

• pnpm exec biome check --write src/backends.ts src/resolve-agent-backend.ts src/resolve-agent-backend.test.ts tests/backends-openai-tools.test.ts
• pnpm exec vitest run src/resolve-agent-backend.test.ts tests/backends-openai-tools.test.ts — 22 tests passed
• pnpm typecheck
• pnpm lint
• pnpm build
• pnpm test — 126 files, 1230 passed, 2 skipped
• pnpm docs:freshness
• pnpm docs:check
• git diff --check
• git fetch origin main && git merge-tree --write-tree origin/main HEAD

[tangletools]

✅ Auto-approved drewstone PR — 6a3b22da

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-07-03T19:42:36Z}