New Kurucz line list parser with straight to HDF5 output

[epassaro]

Trying to get the same DataFrames obtained with the old SQL-based classes.

• Match _get_all_levels_data with AtomData method of the same name
• Match _get_all_lines_data with AtomData method of the same name
• Match levels attribute with AtomData attribute of the same name
• Match lines attribute with AtomData attribute of the same name
• Match levels_prepared match with AtomData attribute of the same name
• Match lines_prepared match with AtomData attribute of the same name

https://nbviewer.jupyter.org/gist/epassaro/82be42e2491511c07bb566354927a780?flush_cache=true

• Add to_hdf method
• Fix for missing ions in gfall.dat
• Figure out issue with He I. Why .drop_duplicates() is not dropping level 3?
• Add docstrings
• Ready to merge

[wkerzendorf]

all tests fail