Implement the new optimization pass for the generated LLVM IR code by making extensive use of virtual registers.

[yomaytk]

This PR implements the new optimization pass for the generated LLVM IR code by making extensive use of virtual registers.
Originally, elfconv (remill) operate the original register computation on the memory by using the State struct that expresses the CPU registers. Instead of, we accelerated the generated LLVM IR speed by possibly using virtual registers not accessing the State struct (i.e. store, load instruction on the LLVM IR).