Handling of 0 and < in nutritional values

Describe the bug

The graphs and stats created by the advanced feature, now treat values with 0 and values with limits (< 0.5) like normal data. This results in strange graphs and stats.

The graph above for example presents the protein values for the category apple nectars. The values around 0.5 and 0.1 are due limits on the package. The zeros ar no actual values either, but rounding errors mad by the producers. Only the values between 0 and 0.1 are actual measurements.

To Reproduce

For instance this graph and stats: https://world.openfoodfacts.org/cgi/search.pl?action=process&tagtype_0=categories&tag_contains_0=contains&tag_0=apple-nectars&sort_by=unique_scans_n&page_size=20&axis_x=energy&axis_y=proteins&graph=1

Expected behavior

The maxima (<) and 0 should be treated as special values:

0 is in fact a rounding error, better is to interpret this as < 0.5. I reality it never can be 0. Some producers note 0.0 (which is now lost in records), which could be interpreted as < 0.05. (at least that is what I learned to do for my physics degree);
exclude the maxima (including the 0) from the stats calculations;
for the graphs have the possibility to exclude the maxima and/or display them in another way (a downward arrow would be best);

Number of products impacted

Any product with 0 or maxima.

Part of

5751

openfoodfacts / openfoodfacts-server