proposed enhancements

[DHekstra]

1. add a weight for know apo data sets in the loss function;
2. add a round of outlier rejection after the first round of VAE training and retrain (I showed an example of Fourier ripples).
3. enable use of the DW extrapolator (Add dw_extrapolator.py rs-station/rs-booster#61, mle_dw_extrapolator addition rs-station/rs-booster#62)
4. establish proper unit tests
5. establish a full CLI
6. ds.dropna().write_mtz(file) in pipeline.ipynb generates a warning for each dataset
7. enable down-weighting of dataset with known hits