Made-up test data for the Calculate RSC and Pairwise difference loss tabs of https://strsc.org/analyze

None of these numbers is a measurement of any tissue. Each gene has a smooth broad pattern and a fine,
spot-to-spot pattern; counts are Poisson draws from these rates (scripts/make_scales_test.py).

spots/     Relative scales. Two sections, 30 genes, spots 100 um apart on a hexagonal lattice.
           true_rates.csv: log(1 + rate), the rates the counts were drawn from.
           broad_only.csv: the same without the fine pattern.
grid/      Physical scales. One section on a 64-um grid, 64 x 64 sites, 30 genes, the same two predictions.
training/  Pairwise difference loss. Three sections, 20 genes, and 48 made-up image features per spot.
cells/     Single cells (layout Cells). One section of 3,000 cells over 600 x 600 um, 30 genes, two classes
           (Neoplastic, Other) in the cell_class column, the same two kinds of prediction (scripts/make_cells_test.py).
