In the past, we've had outages where large test results coming
from DUTs caused load issues at various back-end stages of
processing of results.  A common theme in those outages has
been that we couldn't reliably find example jobs with the
too-large test results.  Without those examples, we were
dependent on luck to find what the large data was, and what
was causing it.

We should calculate the size of all test results, and store that
size someplace convenient like the AFE database.  If we implement
capping of the sort suggested in bugs  716218  and  716271 , the size
in the database should reflect the size of the results before any
cap was applied.