Script t.systest runs the test (building separate graham and robinson
word lists), then runs bogofilter a dozen times on 3 test messages,
using both graham and robinson methods; and compares new test outputs
with saved, known-good results.

Directory t.systest.d/inputs contains the good and spam mailboxes and
3 test messages.  The 3 messages have different bogosities, 1 tests as
ham for both methods, 1 tests as spam for both, and 1 tests
differently.

Directory t.systest.d/outputs containns outputs of 12 bogofilter runs
with the 3 messages - 6 runs each for graham and robinson and with
various levels of verbosity.  Also the ham/spam message is processed
via "bogofilter -s" to increase its spamicity and is run through
bogofilter again - resulting  in spam/spam classification.
