Median Selection Subset Aggregation for Parallel Inference