Just to start off with, I'm not an expert; what I know is from my undergraduate days taking way too many economics classes. If someone is more knowledgeable feel free to call me out.
You're not losing much by skipping farm workers; as I noted above farm payroll changes seasonally, so skipping farm payrolls just helps to smooth the data - I suppose you could also consider it as another way to seasonally-adjust employment figure, although that is stretching the term quite a bit.
In addition farm hands for picking/planting are poorly paid and tend to have vague residency status (not making a political point, just economic reality) and so including that data does not change related figures such as GDP too much. It wouldn't surprise me if politics encourages the reporting of non-farm data due to the issue I mentioned above of vague residency status; see the current DACA fight for how contentious discussions like this can get.
TLDR the numbers work out better and nobody of importance cares about the difference
It's nothing malicious; it's a common way to express employment figures. Farm payrolls tend to swell and contract seasonally (to pick/plant and much less work in-between) so "non-farm workforce" is a way of smoothing out the numbers.
I see quite a few people in this thread blaming the high cost of AWS. I don't understand, why is this a problem with AWS - isn't it a problem with freeloading? Yes, moving hosting may save you some bucks, but fundamentally isn't the problem the large number of freeloaders?
If even 5% of the 6 million users paid $20 a year, Soup would have $6 million a year - more than enough to run a small company on.
Since you're reading this, a feature request: I would love it if you could put up a REST endpoint for extracting all the images (example code here [1]) on a web page, and more endpoints for extracting all the links, script addresses, etc.
I was trying to do that on Browserless but couldn't get the final file download to work (I adapted Stack Overflow code linked below to put all the web pages images into a ZIP file and download that) - presently I'm running this on a Google Cloud Function, which is working but I'd rather outsource it to you, especially since the function chokes on large web pages (possibly it needs more RAM than the 2GB limit currently available in GCF?).
You're not losing much by skipping farm workers; as I noted above farm payroll changes seasonally, so skipping farm payrolls just helps to smooth the data - I suppose you could also consider it as another way to seasonally-adjust employment figure, although that is stretching the term quite a bit.
In addition farm hands for picking/planting are poorly paid and tend to have vague residency status (not making a political point, just economic reality) and so including that data does not change related figures such as GDP too much. It wouldn't surprise me if politics encourages the reporting of non-farm data due to the issue I mentioned above of vague residency status; see the current DACA fight for how contentious discussions like this can get.
TLDR the numbers work out better and nobody of importance cares about the difference