CLIF interpreter: handle divide-by-zero traps

[cfallin]

We received an oss-fuzz bug report yesterday with a very simple test case:

function u0:0(i8) system_v {
    block0(v0: i8):
        v1 = udiv v0, v0
        return
    }

with v0 = 0 as input.

We don't yet handle traps in the differential fuzzing (one step in the checklist in #3050) so this might be an easy starting-point for that issue: we should refactor the interpreter to carry through a Result of some sort such that traps propagate upward, and then reject inputs that trap.

cc @afonso360 @abrown