I generated these as a by-product of using afl-fuzz on the testbn
executable. I'd hoped it would find interesting bugs in the bignum
code itself, but in fact it mostly found bugs in the parsing code in
the test main(). However, even so, it did produce a list of tests
which reach as much of the code as afl-fuzz was able to, and that
seems like a useful thing to save.
Also, of course, just having a place to put explicitly specified
bignum tests is a good thing anyway, because then we can put
regression tests on the end of it every time we fix bugs.
