Add operations compute_dataset() and compute_data_frame()

[forman]

Expected behavior

Cate uses two basic data models, Xarray's Dataset and Pandas' DataFrame (with its GeoPandas extension GeoDataFrame).

Both Xarray and Pandas provide extremely powerful APIs. Cate should provide two operations that would offer using these APIs in a generic way:

1. compute_dataset(ds: Optional[xr.Dataset], script: str) -> xr.Dataset for gridded data
2. compute_data_frame(df: Optional[pd.DataFrame], script: str) -> xr.DataFrame for feature attribute tables / vector data originating from CSV, Excel, Shapefiles, GeoJSON...

In both cases we execute a Python script where either a given ds: xr.Dataset or df: xr.DataFrame provides the namespace to be used in the script. In addition all current workspace resources are included in the namespace. The script is then used to compute new variables of type xr.DataArray or pd.DataSeries using the Xarray API and Pandas API which will be collected to compile a new xr.Dataset or xr.DataFrame respectively.

Actual behavior

We have a prototype computeop which will be renamed to compute_dataset. There is no compute_data_frame yet.

Specifications

Cate 1.0 - 2.0.dev15