Feature Request: Add Kimi K3 open source model as a new channel

## Feature Request

Moonshot AI released **Kimi K3** as open source on July 27, 2026. Since aiproxy already supports multi-tenant AI gateway routing with intelligent channel selection and load balancing, K3 would be a valuable addition to the channel lineup.

## Why K3 makes sense for aiproxy

aiproxy is the Sealos ecosystem's production-grade AI gateway with features like smart retry, priority channel selection, and MCP support. Adding K3 as a channel fits naturally:

- **OpenAI-compatible API**: K3 follows the standard `/v1/chat/completions` protocol, which means it can be added as another OpenAI-compatible channel in aiproxy's plugin-based architecture -- essentially a config entry with the K3 base URL and model name
- **Chinese model coverage**: K3 strengthens aiproxy's coverage of Chinese LLMs, which is important for Sealos ecosystem users who need both domestic and international model access
- **Multi-channel load balancing**: aiproxy's smart retry + priority channel selection means K3 can be mixed into routing pools alongside other models, providing more options for cost optimization and failover
- **Self-hosted scenario**: Since K3 is open source, Sealos users could deploy their own K3 instances and register them as private channels through aiproxy, keeping data within their infrastructure

## Technical implementation

Adding K3 should be straightforward -- it would follow the existing OpenAI-compatible channel registration pattern. The configuration would specify the K3 base URL, API key handling, and model name mappings.

## Context

We recently completed a K3 review and setup via API gateway through TeamoRouter's enterprise platform. The model's OpenAI-compatible interface makes channel integration clean. For aiproxy users in the Sealos ecosystem, having K3 as a supported channel would expand their routing options with a capable open source model.

Start by locating aiproxy's existing OpenAI-compatible channel registration pattern and compare how channel configuration handles base URLs, API keys, and model name mappings. Add Kimi K3 using that pattern, then verify that its OpenAI-compatible chat completions endpoint and model mapping work through the gateway.

Written by the indexing model from the issue text.