fix: LLM 服务去除重复默认值，新增 httpClientTimeout 参数

- 去除 model/endpoint 的 fallback 默认值
- HTTP Client 超时从 config 传入

backend/internal/ai/llm/openai.go

@@ -26,24 +26,23 @@ type OpenAIService struct {
 }
 
 // NewOpenAIService 创建 OpenAI LLM 服务。
-func NewOpenAIService(apiKey, model, endpoint string, timeoutSec int, logger *zap.SugaredLogger) *OpenAIService {
+// model、endpoint 由 config 层保证非空。
-	if model == "" {
+func NewOpenAIService(apiKey, model, endpoint string, timeoutSec, httpClientTimeoutSec int, logger *zap.SugaredLogger) *OpenAIService {
-		model = "gpt-4o"
-	}
-	if endpoint == "" {
-		endpoint = "https://api.openai.com/v1"
-	}
 	timeout := time.Duration(timeoutSec) * time.Second
 	if timeout <= 0 {
 		timeout = 10 * time.Second
 	}
+	httpClientTimeout := time.Duration(httpClientTimeoutSec) * time.Second
+	if httpClientTimeout <= 0 {
+		httpClientTimeout = 60 * time.Second
+	}
 	return &OpenAIService{
 		apiKey:   apiKey,
 		model:    model,
 		endpoint: endpoint,
 		timeout:  timeout,
 		logger:   logger,
-		client:   &http.Client{Timeout: 60 * time.Second}, // HTTP client timeout > LLM timeout
+		client:   &http.Client{Timeout: httpClientTimeout},
 	}
 }