[stub] Scrape and harvest dataset documentation

[nightsh]

Dataset documentation is, in out CKAN extension:

• a package
• has type = documentation
• has a relationship with a dataset type package
• has at least one resource

We can attempt to automatically extract data profile documentation from HTML pages following the rules:

• if a page has at least one resource, extract it as dataset (as we already do)
• if there's a dataset on page, check for document type files (e.g. PDFs)
• from all the document type files, identify which ones to keep (this is the problematic bit, as we might collect many false positives)
• attempt to extract a description for the documentation based on the context of the most relevant document files

In addition to that, we also need to add the resulting documentation to:

• scraper output bucket
  • new documentation subdirectory
• resulting final data.json file
  • new @type value
  • new metadata for Documentation type:
    • dataset - link to the dataset it provides documentation for
• the datajson extension - needs documentation processing capabilities based on the data it finds in the data.json file

Tasks:

• amend the scraper output bucket and the output traversal functions to use the new structure
• implement the new parsing rules to get documentation dumps
• implement the new type in the datajson schema
• add the found documentations to the datajson output

Acceptance criteria

• the scraping process generates (and links) documentation for some of the produced datasets