fuel-posttrain 0.20.0

Post-training model compression for the Fuel ML framework: quantization-calibration and pruning techniques applied to a loaded model's weights.