feat(chat): expose temperature/top_p controls

Context (Phase 3 — future)

Esperanto 2.21.0 added per-call max_tokens / temperature / top_p overrides on chat_complete() / achat_complete() for all LLM providers. Open Notebook currently only threads max_tokens (open_notebook/ai/provision.py); temperature and top_p are not exposed.

Possible scope

• Allow temperature / top_p overrides through provision_langchain_model() / ModelManager.
• Surface controls in chat / model settings UI + i18n.
• Note: for Anthropic, top_p is dropped when temperature is set (provider rule).

Notes

Future / not scheduled. Part of the esperanto 2.22.0 follow-ups.