Kimi Code provider: expose reasoning effort / thinking control (K3 low/high/max, K2.7 thinking toggle)

The Kimi Code provider (kimi, OpenAI-compatible profile, https://api.kimi.com/coding/v1, default model kimi-for-coding) currently offers no way to control reasoning effort in jcode.

Per Kimi's official docs, the current models expose different thinking controls:

• Kimi K3: thinking is always on and accepts a top-level reasoning_effort parameter with three levels: low / high / max (default max). Docs: https://platform.kimi.com/docs/guide/kimi-k3-quickstart
• Kimi K2.7: only supports a thinking on/off toggle, with no effort levels.

It would be great if jcode's effort selection worked for this provider:

• Map jcode effort levels onto K3's reasoning_effort (e.g. low -> low, high -> high, xhigh/max -> max).
• For K2.7, map the effort choice onto the thinking toggle (e.g. none -> thinking disabled, any effort level -> enabled).

Happy to help test. Thanks!

Start by locating the Kimi provider and jcode's effort-selection mapping; the issue identifies the kimi OpenAI-compatible profile and its API endpoint, but no files or tests. Confirm how K3 accepts reasoning_effort and how K2.7 represents its thinking toggle, then verify that each requested effort maps correctly for both model families.