iris.util.new_axis gives odd result if existing dim coord is passed

[rcomer]

🐛 Bug Report

If the scalar coord passed to iris.util.new_axis is already a dimension coordinate, an anonymous new dimension is created.

How To Reproduce

import iris
import iris.cube

print(iris.__version__)

cube = iris.cube.Cube(1)
coord = iris.coords.DimCoord(1, long_name="spam")
cube.add_aux_coord(coord)

new_cube = iris.util.new_axis(cube, coord)
new_new_cube = iris.util.new_axis(new_cube, coord)

for c in [cube, new_cube, new_new_cube]:
    print(c)

output:

3.2.dev0
unknown / (unknown)                 (scalar cube)
    Scalar coordinates:
        spam                        1
unknown / (unknown)                 (spam: 1)
    Dimension coordinates:
        spam                             x
unknown / (unknown)                 (-- : 1; spam: 1)
    Dimension coordinates:
        spam                            -        x

Expected behaviour

I think we should probably get an exception if our "spam" coord is already a dim coord. It might be as simple as adding dim_coords=False in the call to src_cube.coord here:

iris/lib/iris/util.py

Lines 1116 to 1117 in ab1e699

	if scalar_coord is not None:
	scalar_coord = src_cube.coord(scalar_coord)

Environment

• OS & Version: RHEL 7.9
• Iris Version: 3.2.dev0