Sounds like a solution in search of a problem. To machine-read every item's nutrition label, you have to read every item, one at a time. So maybe I can find the food that doesn't contain nuts, but I still have to scan every food item first.

There's an easier way to do that: look for the label on the front that says "nut-free", or the label that says "may contain nuts" or "processed in a facility that processes nuts", or the primary ingredient that is not a nut, etc. No phone and app needed, no encoding data, anyone can do it, it's easy to find. Plus it's not easy to adopt a machine-encoding standard to changes in trends or society (like the difference between "contains nuts" and "processed in a facility that may process nuts"). We also already have OCR that can read all this text, and it's all laid out in a uniform way, so machine processing isn't that hard for existing labels.