thin previous_results submission to only the required attributes _before_ including

[austinorr]

its easier for user to send all previous results than to thin the results per the previous_results_keys response.

suggested fix is for nereid to automatically filter the user-submitted previous results blob for just those keys/columns that should be propagated downstream per the watershed.utils.attrs_to_resubmit function.

it should also be possible for nereid to further thin these to just the node_ids which are 'leafs' of the submitted graph object.

[austinorr]

@NickLacunha I took this note while we were working over the weekend prior to the go-live demo. I think i know how to implement it on my side, but I want to make sure we are on the same page regarding how neptune is building the 'previous_results' blob. I also want to check in on the difference between how results & leaf results are saved between an initial solve on an empty results database and for the delta questions for the top/bottom hourly solves.

This is reminding me that sometimes the dirty nodes are leafs. i know we talked this through at length in the spring, i just want to make sure we are still on the same page.