another limit check bug

[Matthew Fluet]

> Ah, I understand.  I was worried about the C stack frame for the call
> to GC_gc.  But I see that there will just be one, even when switching
> threads.  In the new world, we won't trampoline from one limit check
> loop to another when switching threads.  Sounds good.

Checked in this fix along with -gc-check option for both c-codegen and
x86-codegen.  Remember to remake the runtime before using code generated
by these codegens.