Proof: use GenAI to extract receipt data (instead of basic OCR)

[raphodn]

Story

We already use GenAI (Google Gemini AI) to extract price tag data (see #569)
And we run OCR (Google Cloud Vision) to extract text (see #320), but the data is stored as json.gz and is not used (nor is very easy to leverage, see a full example here)

What about trying to send receipt images to GenAI. With the correct prompt, we could probably get some interesting data out of it !

cc @TTalex & @odin-h