Double empty line between articles.

It does works, however, beside being horribly slow; it has a major issue.

It should separate articles with two empty lines, which it does, sometimes.

`\n\n`

but NOT on all articles.

`\n`

This makes the text output impossible to parse on a large scale, since no consistency is taken, it can happen everywhere.

And I am sure peoples had spent a lot of time figuring it out, as I did, and have end-up to abandon their project completly.

:/

Reproduce the extractor on a Wikipedia dump and inspect output around article boundaries. Trace the formatting path that emits article separators and compare cases showing `\n\n` with those showing `\n`. Done means article boundaries use a consistent separator throughout the output so it can be parsed reliably.

Written by the indexing model from the issue text.