Extraction should fallback on NCDatasets.jl types

[Balinus]

Context
When ClimateTools was designed (3 years ago already!) there was not a lot of options for Geographic types and quite less for climate data. From this perspective, the ClimGrid type was born. However, there is a lack of performance for the extraction of the data from netCDF files but also no support for on-disk datasets (data too large for the available RAM).

Challenge
The challenge will be to align the ClimGrid type with the CFVariable available in NCDatasets.jl. So, either we simply scratch the ClimGrid type found in ClimateBase and replace it with CFVariable (or NCDataset?). There is now some work done in NCDatasets.jl for multiple files dataset and datasets that are too large to fit in RAM and we perhaps should now look into it for a 1.0 release of ClimateTools.

Why
Now that the extraction ecosystem is mature, ClimateTools should now concentrate on climate analysis algorithms (bias correction, etc) and not on the extraction of datasets.

Idea
The API between ClimateTools and NCDatasets.jl should reside in ClimateBase. I'm also posting over there.

Any help, idea would be appreciated!

See also over ClimateTools: JuliaClimate/ClimateTools.jl#119

[Balinus]

See #4 for some discussions.

[Datseris]

Currently we have functions for loading directly from .nc files using NCDatasets.jl.