Unproven Integer Type Checking

[DavePearce]

The goal here is to develop a type check for "unproven" types (e.g. i8 versus i8@prove, etc). Two approaches:

• (Simple) We could use integer range analysis, along with bounded unrolling to given a quick-and-dirty check. This would unroll a given module to some number of rows, then fill in all proven columns with their ranges and then try to deduce the rest. I don't know whether this will work well, however, for modules with CT_MAX. The problem being that these modules have variable rows per call, and this will interfere. We could hard-code knowledge of CT_MAX.
• (Non-inductive Proof) We can improve upon the above by using a simple decision procedure for inequalities and, again, unrolling to some depth.
• (Inductive Proof) Perhaps the hardest, but also the most powerful.