feat(core): allow agents to opt into a small model for lightweight turns

[everest-an]

Allow agents to use a small/fast model for lightweight turns

Problem

The SessionRunner resolves a single model per session (SessionRunnerModel.resolve) and uses it for every provider turn. The Catalog.model.small() helper already exists and is used for title generation, but agent turns never take advantage of it — so lightweight steps (status updates, confirmations, plain-text continuations) pay the full cost of the main model, which slows down long-running multi-step tasks.

Proposal

Add an opt-in small flag to the agent schema (AgentV2.Info). When an agent sets small: true, the runner resolves a small model via the existing Catalog.model.small(providerID) and prefers it for the turn, falling back to the session model when no small model is available.

• Zero behavior change for existing agents (flag defaults to absent).
• Never fails a turn when a small model is missing.
• Reuses the existing Catalog.model.small() (keyword + cost/age scoring) rather than inventing new model-selection logic.

Changes

• packages/schema/src/agent.ts: add optional small boolean to AgentV2.Info
• packages/core/src/session/runner/model.ts: add SessionRunnerModel.resolveSmall
• packages/core/src/session/runner/llm.ts: prefer the small model when agent.info?.small is set
• packages/core/test/session-runner-model.test.ts: unit tests for resolveSmall injection + fallback

Notes

• This is a first slice; a follow-up could add automatic "lightweight turn" detection instead of requiring the explicit flag.

[github-actions]

This PR doesn't fully meet our contributing guidelines and PR template.

What needs to be fixed:

• PR description is missing required template sections. Please use the PR template.

Please edit this PR description to address the above within 2 hours, or it will be automatically closed.

If you believe this was flagged incorrectly, please let a maintainer know.