IntInf asserts

[Stephen Weeks]

> If this is final, could you mail me what change you want and I will do the
> C routines. 

I would like to see it happen, but it's not near the top of my todo.
I'm not sure where it is on Matthew's.

> Also at some point we should talk about an improved interface
> which would speed up IntInf ops in the case where nothing overflows.  I.e.,
> if both args are small (31 bits) and the product is small then it would be
> nice for it not to have to go to C code.  The code to do this is easy if I
> can count on Overflow getting raised, but that wouldn't work in the C backend.
> Also it is probably better to have a little bit of assembler to give me the
> double integer result in that case.

You can count on Overflow these days, in both backends.