The following Pull Request:

• Closes Enable consolidate_metadata check for named dimensions #506. It adds another parameter that enables to check for maps in the case of named dimensions.

This addresses a performance issue when accessing via xarray+pydap ocean velocity from OSCAR. For that dataset latitude and longitude are named, whereas lat and lon respectively have the values (meaning one can request dap responses). In this scenario, xarray requests lat and lon instead of latitude and longitude. The requests grow as 2*N, N being the number of URLS.

This PR then enables caching for lat, lon or any other similar Map, when the set_maps is set to True (False by default).

Example

urls = ['dap4://opendap.earthdata.nasa.gov/collections/C2098858642-POCLOUD/granules/oscar_currents_final_20190816',
 'dap4://opendap.earthdata.nasa.gov/collections/C2098858642-POCLOUD/granules/oscar_currents_final_20190823', 
...
]
%%time
consolidate_metadata(urls, concat_dim='time', safe_mode=False, set_maps=True, session=my_session)

For 100 urls , consolidate_metadata takes ~40 secs. With that, the following takes 1 second:

Notes

1. That the coordinates (e.g. maps) and dimensions do not match spatially.
2. It is possible to bypass consolidate_metadata and create the xarray dataset directly. However, this is ~ 10x slower.
3. There will be some need for more testing, but it works for my purpose