ARBRES extraction v2

TODO

• Identify the examples with different number of columns for the text and for the gloss (for normalisation). And try to columnize stuff?
• Ignore all phonetic stuff.
• For pages with multiple categories, look into lefff, if there's still several alternatives, panic.

Notes

• [[a]] is the lemma of a preposition, the current extraction mistakenly uses it for rannigoù.