fix: increase LLM client timeout to 5 minutes

GPT-5-mini timed out on larger diffs (2min was too short).
LLM calls for code review with full file context can take 2-4min.

llm/client.go

@@ -27,7 +27,7 @@ func NewClient(baseURL, apiKey, model string) *Client {
 		baseURL: strings.TrimRight(baseURL, "/"),
 		apiKey:  apiKey,
 		model:   model,
-		http:    &http.Client{Timeout: 2 * time.Minute},
+		http:    &http.Client{Timeout: 5 * time.Minute},
 	}
 }