test_overflow: macrofy some more, do more tests for free
Obviously a+b==b+a and a*b==b*a, but the implementation of the fallback
checks are not entirely symmetric in how they treat a and b. So we might
as well check the (b,a,r,of) tuple as well as the (a,b,r,of) one for +
and *. Rather than more copy-paste, factor out the common part to
check_one_op.