create intermediate data sets that are usable for intermediate explorations of performance

[elray1]

these would be similar to the data sets in issues #8 #9 and #10, but would be created only on an ongoing basis (at least to start), and each week we would create files for all of the recent reference dates when predictions were made but for which we haven't yet reached the "90 days out" threshold for creating the final evaluation data.