Request Limits vs Token Limits: Which Flat-Rate LLM API Pricing Is Cheaper for Your Workload?