Support checkpointing on IndexSolver #58
Description
The IndexSolver is currently a wrapper around the GeneralSolver. It solves using the GeneralSolver in one go, meaning it solves everything in-memory and then writes everything in a file. Obviously this is not optimal.
Let's start by optimizing the "one go" part. Modify the IndexSolver to support checkpointing. More specifically, we want the IndexSolver to store its state (perhaps through pickling) after fulfilling some criteria (decided by the implementer). What's more, the IndexSolver should be able to recover its state from its last checkpoint and continue solving.
When recovering from the last checkpoint, it's ok to load the entire file into memory. That'll be fixed in a later issue.