Added support for timescaledb

[gedemagt]

Currently the implementation of postgres fails when trying to recover a TimescaleDB, due to their special hypertables.

This PR is mostly a copy of the postgres backend, but with the important difference that the we drop the extension before we drop the tables:

    cursor.execute('''
            DROP EXTENSION IF EXISTS timescaledb CASCADE;
        ''')

[coveralls]

Coverage decreased (-2.5%) to 91.579% when pulling 3021ded on gedemagt:master into 9ad2b3e on TeamHG-Memex:master.

[mehaase]

Thank you for the PR @gedemagt, and I apologize for not responding yet. I'll try to review it soon. I don't have any experience with Timescale DB but this basically looks good. Before I merge, I'll need automated tests to cover the new code and updated docs. If you are able to include those in your PR, that will definitely decrease the time to merge.

This PR is mostly a copy of the postgres backend, but with the important difference that the we drop the extension before we drop the tables:

My main question for review is if the difference between timescale and postgres is very small, can you subclass the postgres class and override one or two methods to support timescale?

[gedemagt]

Good point - I changed it to subclass.

I will have a look at the test as soon as possible.