feat: remove parallel_tool_calls parameter from litellm provider (#10274)

Co-authored-by: Roo Code <[email protected]>

src/api/providers/lite-llm.ts

@@ -133,7 +133,6 @@ export class LiteLLMHandler extends RouterProvider implements SingleCompletionHa
 			},
 			...(useNativeTools && { tools: this.convertToolsForOpenAI(metadata.tools) }),
 			...(useNativeTools && metadata.tool_choice && { tool_choice: metadata.tool_choice }),
-			...(useNativeTools && { parallel_tool_calls: metadata?.parallelToolCalls ?? false }),
 		}
 
 		// GPT-5 models require max_completion_tokens instead of the deprecated max_tokens parameter