Potential improvements to Everest keyword reference

[oyvindeide]

Currently we are using json_shema_for_humans to generate the keyword reference for Everest: https://github.com/equinor/ert/blob/122dae5a7d924be3f4d8d75da662911278ab04ac/docs/everest/conf.py#L51. This has some advantages:

• Always up to date configuration
• Examples available
• Follows json-schema standard, so everything is handled

However it also has some drawbacks:

• Not easy to configure
• Readability is not the best, especially when keywords are optionally None (null)

If possible, we would like to avoid a custom implementation of this, and rely on some existing tool.