feat: support comma-separated AI_MODEL for quick multi-model setup (#870)

Users expected setting AI_MODEL to a comma-separated list to expose
multiple models in the picker, but the value was used verbatim as a
single model id, leaving the picker with only the "Server Default"
fallback.

Add a third-priority fallback in loadEnvServerModelsConfig: when
AI_MODELS_CONFIG and ai-models.json are both absent, AI_MODEL contains
a comma, and AI_PROVIDER is set to a known provider, synthesize an
equivalent ServerModelsConfig with the provider's models trimmed,
deduplicated, and the first marked as default.

Also makes getAIModel and getValidationModel pick the first comma-split
value when falling back to AI_MODEL, so requests started before the
client picker hydrates still resolve to a real model id.

Docs (en/cn/ja) and env.example updated; tests cover the new fallback
plus the no-comma / no-AI_PROVIDER negative cases.

docs/cn/README_CN.md

@@ -217,7 +217,7 @@ npm run dev
 
 ### 服务端多模型配置
 
-管理员可以配置多个服务端模型，让所有用户无需提供个人 API Key 即可使用。通过 `AI_MODELS_CONFIG` 环境变量（JSON 字符串）或 `ai-models.json` 文件配置。
+管理员可以配置多个服务端模型，让所有用户无需提供个人 API Key 即可使用。通过 `AI_MODELS_CONFIG` 环境变量（JSON 字符串）或 `ai-models.json` 文件配置。如果只需要单 provider 下的多个模型，也可以直接在 `AI_MODEL` 中用逗号分隔模型 ID。
 
 **模型要求**：此任务需要强大的模型能力，因为它涉及生成具有严格格式约束的长文本（draw.io XML）。推荐使用 Claude Sonnet 4.5、GPT-5.1、Gemini 3 Pro 和 DeepSeek V3.2/R1。