various updates

[sfluegel05]

Various updates for the new ensemble version. This includes

• support for ensemble update  python-chebifier#32
• better ensemble settings Advanced ensemble settings #15
• a precision/recall setting
• a data disclaimer Disclaimer about data storage #21
• rdkit.js loading twice RDKit.js is loaded twice #26
• inchi support Input via IUPAC names / InChI #6
• Users cannot give feedback to predictions #7
• track usage statistics #25

Todo

• The API changed. Check with users if this breaks anything

[sfluegel05]

We decided not to include statistics about predicted molecules - while we would only store the number of predictions in an aggregated form, user concerns for privacy are more important

[sfluegel05]

more todos for alignment with python-chebifier

• replace .json config with .yml config (from python-chebifier) for consistency (mostly - the description texts stay
• check if the file download uses the same format as python-chebifier
• replace custom WMV-F1 extension with python-chebifier version of model weights

[sfluegel05]

more todos for alignment with python-chebifier

* [x]  replace .json config with .yml config (from python-chebifier) for consistency (mostly - the description texts stay

* [x]  check if the file download uses the same format as python-chebifier

* [x]  replace custom WMV-F1 extension with python-chebifier version of model weights

These todos are now cleared. I also tried to move as much logic as possible to python-chebifier, in particular the caching of the prediction smoother and the chebi graph.