build
Build PrecomputedData from primitive AP rows.
Input is a mapping systemId -> DoubleArray(AP per topic).
Steps
Copy rows into a dense row-major matrix
[S][T]Compute full-set means per system
Build column views
[T][S]for fast ± updates
Return
a PrecomputedData structure containing row-major, column-major, and mean views.
Parameters
average
Mapping from system ID to per-topic AP scores. Each entry is a DoubleArray where array[t] is the AP of the system on topic t.
Throws
if input is empty or inconsistent.