more semantic data structure of rdf input

[koo5]

There are at least two places where things aren't fully semantical:

1. a request does not reference data of individual sheets semantically, instead, a request references a list of sheets, and it's up to the receiving end to walk the sheets and extract references to root data items

unit_valueses is an example of a sheet with higher complexity than can be captured declaratively by our rdf templates-
it has to contain "custom" columns, where each column has a header containing a date, and the values in the cells are related to that date.

The plan is to add cases in our prolog code to use data that are expressed semantically. At least for unit_valueses. But even for sheets, the change is fairly trivial, rather than having the python sdk code assert that request has sheets, and one of those sheets has a data, we directly assert that request has data. A lot of the prolog code actually still expects this format from before (!?)