Document the significance of negative values for count-from-zero observations

[mikaelhg]

The airquality_urban producer has a bunch of negative values polluting its PM10_PT1H_avg hourly aggregation, and those seem to randomly pull the average to the negative, and while I don't any direct evidence of this, even when the average doesn't go negative, those random negative values probably induce a systematic error to the data available to the consumers of the endpoint.

It seems pretty obvious, that when you count up from zero, you can never go negative.

Please fix the averaging query so that it doesn't include negative values at all when it calculates the average for an hour.

... and please fix the sensors and the integration between the sensors and the master data, to handle errors and uncertainties better.