Convert vertices to bounds for nD arrays

Start with cf_xarray/helpers.py and the vertices_to_bounds entry point shown in the traceback. Reproduce the supplied nD DataArray example and compare it with the working single-profile call. Done means the full array with x, y, time, and test dimensions is accepted and converted into the requested bnds and test bounds.

Thank you for developing this amazing package, I am only slowly appreciating its full amazingness.

I am currently trying to convert an array of layer depths that varies in time and space, which represent cell vertices in the vertical into a bound style array.

This is my original array lev_vertices:

For a minimal example you can create a similar array with:

lev_vertices = xr.DataArray(np.arange(9), dims='test').expand_dims(x=2,y=2, time=2)
lev_vertices

I would like to convert the dimension test into bounds.
This works as expected when operating on a single profile:

cf_xarray.vertices_to_bounds(
    lev_bounds.isel(time=0,x=0, y=0).load(), out_dims=["bnds", "test"]
)

But I cant figure out how to make this work on the full array...

cf_xarray.vertices_to_bounds(
    lev_vertices.load(), out_dims=["bnds", "test"]
)

gives

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-76-4487305bbcdc> in <module>
----> 1 cf_xarray.vertices_to_bounds(
      2     lev_vertices.load(), out_dims=["bnds", "test"]
      3 )

/scratch/gpfs/jbusecke/conda_tigercpu/envs/aguadv_omz_busecke_2021/lib/python3.8/site-packages/cf_xarray/helpers.py in vertices_to_bounds(vertices, out_dims)
    116         )
    117     else:
--> 118         raise ValueError(
    119             f"vertices format not understood. Got {vertices.dims} with shape {vertices.shape}."
    120         )

ValueError: vertices format not understood. Got ('x', 'y', 'time', 'test') with shape (2, 2, 2, 9).

I am wondering if I am missing an option to ignore dimensions here? I was thinking of using xr.map_blocks, but that would impose restrictions on my chunking (which might affect performance). So I first wanted to check in here, and see if there is an easy fix.