Convenience function for making a regridding target Cube of specified shape

[trexfeathers]

✨ Feature Request

• As a: user of Iris regridding
• I want: an easier way to create a target Cube from scratch, with a specified shape
• So that: I can easily regrid my source data even when I do not have a target that includes data

A function in iris.util that returns a Cube would seem like the simplest implementation, with various config for coordinate lengths, coordinate units, etcetera. Ideally we could not only create standard rectilinear space (1-dimensional coordinates) but also curvilinear (2-dimensional coordinates) and mesh (using MeshCoords).

To advertise this more strongly we could consider allowing regridders to accept a 'cube shape specification' as the target, in addition to accepting an actual Cube. Perhaps by introducing a CubeShape class somewhere?

Motivation

Currently regridding without a loaded target Cube requires creating a Cube from scratch, with all the necessary Coords attached. Many users are unaware this is possible, and many more find it difficult. More significant: it is not obvious to a non-developer that Iris regridding will work fine with a Cube that has blank data.

Additional context

Mentioned in #4447. Any convenience should be revisited if we ever get the ability to create a data-less Cube.

A private testing convenience exists for this in esmf-regrid.tests.unit.schemes.test__cube_to_GridInfo._grid_cube()

Expand for some typical code that users currently need for this case:

min_x = -20
max_x = 20
min_y = 40
max_y = 60

points_x = np.linspace(min_x, max_x, 100)
points_y = np.linspace(min_y, max_y, 100)

my_cs = GeogCS(DEFAULT_SPHERICAL_EARTH_RADIUS)

coord_x = DimCoord(points_x, standard_name="longitude", units="degrees", coord_system=my_cs)
coord_y = DimCoord(points_y, standard_name="latitude", units="degrees", coord_system=my_cs)
target_cube = Cube(np.zeros((100, 100)), dim_coords_and_dims=[(coord_x, 0), (coord_y, 1)])

regridded_cube = my_cube.regrid(target_cube, Nearest())