Feature parity with xarray

[rafaqz]

With #168 this is surprisingly close, besides all the things that GeoData does that xarray can't do.

But clearly missing are:

• [ ] Dask-like processing for larger-than-memory files. This should already work in a simple way by broadcasting from one open file to another - the file can be larger than ram. But its not parallel or distributed. I am unlikely to write this unless someone specifically needs it, as I never work with single files larger than ram. But maybe Dagger.jl can help here?
• [ ] netcdf layer-specific compression settings.
• [x] easily adding new layers to an existing stack
• [x] split apply combine ?

Also surprisingly, most of the "technical vision" for xarray is already possible in GeoData.jl due to composability of julia packages.

Battle hardening is probably one of the main things missing :laughing:

[felixcremer]

For the first point we can use YAXArrays.jl or later on DiskArrayEngine.jl . Especially, after we finished https://github.com/JuliaDataCubes/YAXArrays.jl/pull/249 so that these two packages are playing more nicely together.