Draft for AbstractReader.py.

This version is just a draft open for discussion.

Some of the changes: Removed the get_database method. This would be purely implementation related… If one needs to do something else than extract data then it should create it's own connector, I don't understand the need to expose this. Created an enum for the aggregation functions. These are the one we support in timescaledb. I think it should contain all the aggregation functions that are supported in any of our backends. All custom aggregation should then be shared and not be linked to the backend. Removed the correlated argument, for more or less the same reason, this can and should be shared between all the backends. I am actually interested in knowing if we could do this at the DB level, I believe we can, so we might want it back. Added some documentation on the expected format and so on. I have no idea about the performances of the underlying containers, we might want to change it (some suggested the use of pandas).