Semiramis - harmonize old and new examples

Currently we have two examples for Semiramis:

1. the old parametric model, 470k dataset and masking
2. the new parametric model, without masking - for which we provide an example a) "full pipeline" and b) "from dataset" with 10k samples

We could harmonize them to have a more elegant structure and to be able to highlight the difference better.

• create one common dataset for 1) and 2b)
• harmonize old and new notebooks
• #120 (closed)
• readme