Cherry-Picking in Time Series Forecasting: How to Select Datasets to Make Your Model Shine