Evaluating Large Language Models for Workload Mapping and Scheduling in Heterogeneous HPC Systems