Data transfers tracking + migrations

[mensfeld]

ref #199

This PR also introduces:

• the aggregated states migrations, so I can easily ensure states consistency and so I do not have to worry about cases when some of the data would be missing partially and would only start to appear later. I can just migrate missing stuff.
• It introduces aggregates states version collision detection, allowing us to stop processing and signal if a consumer running older version of karafka would try to work with newer data. We had this for the processes reports, but now it expands to the rest of the data.