LLVM memory model code reorganization

[brianhuffman]

The structure and operation of the LLVM memory model is complicated and hard to understand. We should refactor and reorganize the code to make it easier to figure out.

It would be very helpful if the code was separated into modules in such a way that each module could be understood separately. (The current division into modules Common.hs and Generic.hs is a relic of an older implementation, and doesn't make sense anymore.) In particular, it would be nice to be able to understand the offset-related parts of the memory model separately from the allocation-unit-related parts.

Those modules could also benefit from additional documentation.

[langston-barrett]

It would also be nice to design a more abstract interface for the memory model (e.g., a typeclass) so that client code could more easily switch between different implementations (e.g., a bump-allocator based implementation based on SMT arrays).