Improve CSV importers

[mauromsl]

Current csv importers could use some of the following improvements:

• Schema validation: rather than directly unpacking the csv lines we could have some sort of schema to compare the structure against, and return meaningful errors in case of schema validation.
• Allow users to download a template with the correct headers in .xls or .csv formats
• Rewrite the html form with a Django forms.Form for the file upload and add a field to specify csv delimiter as well as header separator.
• Abstract the validation an import so it works for all kinds of importers. Then each importer only has to provide a schema and the logic to process the csv

[mauromsl]

We have discussed a v1 implementation for this issue that would involve:

• Moving the validation logic used in Janeway core to callables/functions that can be invoked by importers. This will allow the same logic to be re-used and avoid traps where logic changes across import routines
• Update the CSV importer to invoke those functions programatically on a per field basis
• Wrap article processing in an atomic transaction to ensure problems ingesting a row don't leak to the rest of the import procedure.
• When errors occur, return a generated CSV with just the faulty rows that have been rolled back.

[mauromsl]

Note: Mauro to raise an issue in core and make this depend on it