Try harder OCR detection for some labels in case of OCR typos

[teolemon]

ACRICVLTURE\nSIOLOGIQUE\n1\nCERTIFIÉ PAR FR-BID-T0

https://static.openfoodfacts.org/images/products/376/012/121/3259/1.json

[raphael0202]

I'm not sure the response lies on the regex side, more on the OCR side (we may run the OCR on the same image once Google Cloud release a new version of its model)