Hydpy

hstream_v1` calculated `nan` values for some configurations. This should now be fixed.

We now perform proper integration testing for `hstream_v1`, so that this problem should not recur.

Major improvements:
* automatic writing of averaged time series (see module "variabletools")
* reading and writing of NetCDF files (see module "netcdftools")
* configuration of workflows with XML files (see module "xmltools")
* command line support (see module "hyd.py")
* interoperability with Delf-FEWS
* simplified installation using "pip" on 64-bit Windows
* runnable on Python 3.6 and 3.7 (but not on Python 2.7 anymore)
* Windows based testing on AppVeyor

Some of the more important features currently under development:

* Temporal smoothing of potential evaporation for HydPy-L-Land.
* Allow negative discharge of HydPy-L-Land for subbasins with large water areas under dry conditions (optionally).
* Add the `RestrictTargetedRelease` option to HydPy-D.
* Improve some error handling functionalities.

Major fixes:
* Method `save_allseries` of class `Nodes` does now work properly.
* Method `init_models` can now handle the case of missing control files.

Minor fixes:
* The handling of missing directories should now be more consistent.

Major fixes:

* The calculated meltwater of the glacier itself was overwritten by the meltwater of its covering snow layer. We fixed this error.

Minor fixes:

* The adaption of precipitation and potential evaporation to altitude cannot result in negative values anymore.

Major fixes:

* When implementing the two additional water area types SEE and FLUSS into HydPy-L-Land, an error had crept in the calculation of direct discharge from sealed surfaces (VERS). This error is now fixed.

Minor fixes:

* Some information on the status of different methods for handling time series data had been printed on the screen multiple times. This error is now fixed.