Do not duplicate harvest_extras if exist in root schema

[pdelboca]

This PR refactors and document the logic of our dataset_before_index method.

With the new changes, harvest metadata will not be added to extras if it exist in the root schema of the dataset. This will allow other extensions or implementations to add harvest metadata to the main package schema without getting duplicated errors when trying to update or patch the dataset.

[pdelboca]

@amercader As a continuation of this PR I want to add some documentation about pkg_dict, data_dict and validated_data_dict on CKAN Docs.

I still don't get quite well the uses and difference betwen them.

[amercader]

@pdelboca This looks good!

re the different dicts:

• pkg_dict means "the dataset dict that will be sent to Solr for indexing"
• data_dict means the default dataset metadata dict without any validation applied (eg all custom fields are extras)
• validated_data_dict means the dataset metadata dict after applying validation (eg what you get on package_show with the schema customizations applied)