✨ Feature Request

It should be possible to use this code when loading a NetCDF file, and only these 3 Cubes should be created. I.e. loading a file with 60 variables should not create 60 Cubes before constraining:

iris.load_cubes(file, ['var1', 'var2, 'var3'])

See #6223

Motivation

Currently NetCDF files with many variables are very slow to load, even if you only want a few of the variables (i.e. Cubes) to load. This is because all the Cubes must be created first before filtering can be done.