Test exptar_*_to_buffer results instead of full chain

[jonasjucker]

Currently we test against the result of the final extpar-file after the consistency_check.f90.
This takes a lot of time and many configurations are tested multiple times.

Switchting to a more fine grained testing approach based on the intermediate buffer files would it make much easier to
test more configurations.
Furthermore one could parallelize testing even more and submit each integration test for a buffer exe individually.

Also one could test much easier different namelist combinations for the same script, because execution time would be in the order of seconds, not teens of minutes.

Also one could consider moving to pytest entirely, since it offers more flexibility with regard to test design.
The testsuite currently involves quite some bash-scripting.