Hdf5 and database restructuring

[philippkraft]

Finally a hdf5 datawriter is available based on pytables. To keep pytables an optional dependency, all datawriters got their own module inside of the new package database. I applied some magic in database.init.py to allow for some syntactic sugar with Python 3.7+, but this should not create harm in older versions. Might get kicked out if tests fail on Travis

Side effects

During writing the tests, I found a strange use of the input function as a reaction on IOErrors in the database csv and sql. That feature is harmful for high performance computing applications or other usages beyond the standard command line and even there, this is not expected behaviour of a libraray. This has been removed (d6a4aeb).

[philippkraft]

Ok, pytables is missing in the dependencies

[philippkraft]

HDF5 works and can be merged from my part. @thouska, can you check the compatibilty with the analyzer? You can load the hdf5 table into a recarray with the following code:

with tables.open_file('dbname.h5') as db:
    return db.root[dbname][:]

The dtype differs from the load_csv function in the analyzer only for the simulation.