Msbuddy

- Update deep learning features & models for molecular formula annotation.
Now we use a learning-to-rank model to rank the formula candidates.
- Update MS2 preprocessing: reserve top 6 peaks (user-defined) for each 50 Da window.
- Update database source: now we use the GitHub release to manage the database source.
- Update candidate formula search, FDR estimation, ML etc.

- Bug fix for load_mgf (now can load metabolite features with MS1 and MS2 spectra using the same identifier).
- Update command line version.

- Update base class "Adduct".

- Update MsbuddyConfig class and PreprocessedMS2 class.

- Bug fixes for export results in cmd line version.
- Add version control for downloaded data (loaded DBs).

- Update export: pd.Dataframe 'append' deprecated.