add: structured json dataloader

[iwilltry42]

Example format:

{
  "metadata": {
    "source": "https://example.com",
    "filename": "foo.pdf"
  },
  "documents": [
    {
      "metadata": {
        "page": 1
      },
        "content": "This is the first page of the document."
    },
    {
      "metadata": {
        "page": 2
      },
      "content": "This is the second page of the document."
    }
  ]
}

knowledge load supports this as a default now: knowledge load mydoc.pdf - will print the loaded content from mydoc.pdf in the above format.
You may as well verify your own structured output via diff <(cat examples/structured-ingestion/example.json) <(knowledge load examples/structured-ingestion/example.json --loader "structured" -) (there may be formatting differences).
Note: knowledge load sets the global source metadata to the source file that you give it.