[Docs] Clarify Kimi Code model reasoning behavior

[zoomote]

​Opened on behalf of @taltas. Follow up by mentioning @roomote, in the web UI, or in Discord.

What changed

Corrected the existing Kimi Code provider guide to document all four current model IDs and their actual request behavior:

• K3 models use configurable low/high/max reasoning effort with high as the default.
• K2.7 Code models keep thinking enabled and preserve reasoning across turns.
• Direct Kimi Code keeps a 131K output default instead of inheriting the 32K OpenRouter cap.

The page also clarifies OAuth/API-key setup, membership-dependent context limits, model-switching guidance, and the distinction between Kimi Code and the pay-as-you-go Moonshot provider.

Why this change was made

The previous page described only K3, listed the wrong 32,768-token output limit, and treated max as the default effort. This update aligns the guide with the direct provider behavior implemented in Zoo-Code PR #1341 and explicitly follows the OpenRouter-specific work in Zoo-Code PR #1325.

Impact

Kimi Code users can now choose a model with accurate expectations about context, output limits, reasoning controls, and preserved thinking.

Related PRs

• Zoo-Code provider behavior: [Fix] Kimi Code sends incorrect reasoning parameters for subscription models Zoo-Code#1341
• OpenRouter Kimi K3 fix: fix(openrouter): profile moonshotai/kimi-k3 with truthful max_tokens and reasoning effort Zoo-Code#1325

Screenshots