related to #8, #33:

Finally found a way to use the controlled vocabulary in pytests without
hardcoding them instead of getting them from the database itself.
BUT: At several places, 'import toardb' had to be inserted
     (to avoid circular imports when importing only the controlled vocabulary)
AND: pytest would only work when stationmeta.models_annotation
     imports generic.models
     (This seems to be related to the order in which tables are created.)

Deleted all '# coding: utf-8' lines from source code,
since this is standard with python3.