Pydda

This fixes bugs in the calculation of the weights from the grid mask and also fixes the functionality behind throwing a warning for missing cartopy package.

I am proud to announce a new release of PyDDA. This release adds:
* A bug fix for the calculation of the radial velocity cost function to be more consistent with MultiDop. This will change the coefficients you use for this constraint.
* Improved the capability to adjust the thickness and sizes of quivers and streamline arrows for better visualizations.
* Support for integrating ERA-Interim data into retrieval. This feature can even automatically download and interpolate the ERA-Interim data to the analysis grid, but will require the ECMWF Web API to be installed in order to do so.

The ECMWF Web API, an optional dependency of PyDDA, can be downloaded here:

https://confluence.ecmwf.int/display/WEBAPI/Access+ECMWF+Public+Datasets

One will need to follow the instructions at the above link to set up an account with ECMWF and to use this account with the Web API.

This is the same as 0.3.0, but with a fix to the MANIFEST.in file for packaging.

This new release fixes a few bugs with the visualization module as well as adds the following features:
* Use of a 2-pass nested retrieval for running PyDDA with very large (1000's of points in x and y) grids. This will use a distributed cluster to map the problem of optimization to smaller chunks distributed onto workers.
* Quiver plots

There is now a dependency on dask and distributed for the nested retrieval.

This is the second release of PyDDA. This release supports:

1. Use of HRRR, WRF, and radar data as observational/model constraints in retrieval.
An arbitrary number of radars can be used, with custom weight fields used for each radar.

2. Plotting of gridded data using wind barbs and streamlines.