The code is now using a constraint representation for Jump bounds; an
attempt to reduce the need for strided intervals.
It's a bit unclear how much of a win this is; the jump bounds should
recognize the size of a jump table with fewer needs to repeat as
strided intervals may. However, it doesn't solve the incompleteness
problem that I hoped it would -- if the intra-procedural control flow
graph contains missing edges, the result may still be incomplete.
This code also introduces ParsedBlocks, a basic block representation
that used the ParseTermStmt. A future patch will start including this
in the discoveryinfo, so that later procedures do not need to call
"classifyBlock".