Adaptive Workload Distribution for Accuracy-aware DNN Inference on Collaborative Edge Platforms