Offer plugins a way to add serialized data into the database

We should offer a plugin a callback or some other mechanism to add data into the database during the serialization process. This allows them to more efficient store and retrieve analysis related data.

A potential use case is like a plugin deals control-flow flattening, and it works by first analyzing the function, figuring out the mapping relationship of the basic blocks, and then use a workflow to rewrite the IL. Since the analysis and calculation can be expensive, so it cannot be redo every time the workflow runs, it must be somehow cached. And this cached information needs to be saved into the database as well. And the ability to write data into the database during the serialization process gives the plugin the same capacity to persist the data as the analysis done in the core

An alternative to this is to support function-level metadata (right now we only support binary view level), but I believe that is inferior to directly allow a plugin to do serialization

No files, tests, or entry points are named. Start by reading the existing serialization and binary-view metadata mechanisms, then determine how a plugin could persist analysis data during serialization; done means a documented, usable plugin mechanism for storing and retrieving that data.