feat: new openfoodfacts configuration

[alexgarel]

• index all the fields we want to be able to use
• update API to fetch document to last version

[raphael0202]

A docstring summarizing the transformations would be nice.

[stephanegigandet]

Maybe we should change the schema in product opener instead of here, and use a more descriptive format (e.g. as you did, using hash keys to indicate what it is instead of arrays with different meaning according to the position in the array...)

[alexgarel]

@stephanegigandet we want to deploy quickly.

I think transforming nova_group_markers in the API is a good idea, and we can drop this code after that.

But for images, I think using image id as hash key is ok for the API (and easy to manipulate in JS), it's just that for ES we want a list of dicts, but it's more related to the way ES works. So it's ok to transform the data in this scenario.

And this leaves a question to you @raphael0202 and @stephanegigandet: should we transform data back to the hashmap style when we send back results ?

[raphael0202]

And this leaves a question to you @raphael0202 and @stephanegigandet: should we transform data back to the hashmap style when we send back results ?

I think we should stay close to the OFF API, so I would prefer transforming it back to the format returned by Product Opener in v3.4.

[raphael0202]

A few remarks, but otherwise looks good to me!

[alexgarel]

For information, the state of this PR is the following:
We wanted to handle taxonomy synonyms using the synonyms filter of ES but this leads to something like a mapping explosion, that is the startup time of ES is really really long (>40 min).

• A discussion about that is happening here: How should we handle taxonomies synonyms #333