`DataArray.drop_sel()` along a MultiIndex-ed dimension drops too much data

What happened?

Calling the drop_sel method on a DataArray object produces an incorrect result if the axis involved is indexed by a MultiIndex.

Assuming that dimension 'direction' is indexed by a MultiIndex with named levels 'azimuth' and 'elevation' one can attempt to drop a single item along this dimension in two ways:

• call da.drop_sel(direction=(0.0, 0.0)), but this drops all items whose 'azimuth'==0, regardless of 'elevation'
• call da.drop_sel(azimuth=0.0, elevation=0.0), but this results in an error ValueError: dimension 'azimuth' does not have coordinate labels

What did you expect to happen?

I expected that at least the first call would result in a correctly dropped item along dimension 'direction'. This expectation is based on the fact that calling .sel(direction=(0.0, 0.0)) works and returns the element where both 'azimuth' and 'elevation' are 0.0.

It would be convenient for the second call to also work, but the analogous call to .sel(azimuth=0.0, elevation=0.0) does not work either, so I guess this is unsupported.

Minimal Complete Verifiable Example

import numpy as np
import xarray as xr

# Create a DataArray with a MultiIndex-ed dimension:
da = xr.DataArray(
    np.reshape(np.arange(16), (4, 4)),
    dims = ['azimuth', 'elevation'],
    coords={
        'azimuth': range(4),
        'elevation': range(4),
    }).stack(direction=('azimuth', 'elevation'))

# Correctly selects a single item along 'direction'
da.sel(direction=(0, 0))

# Incorrectly removes all items along 'direction' where azimuth==0...
da.drop_sel(direction=(0, 0))

# ...but it should remove just one item, like this call:
da.drop_isel(direction=0)

# This fails with 'ValueError: dimension 'azimuth' does not have coordinate labels':
da.drop_sel(azimuth=0, elevation=0)

Relevant log output

No response

Anything else we need to know?

No response

Environment

INSTALLED VERSIONS

commit: None
python: 3.9.7 (default, Sep 16 2021, 13:09:58)
[GCC 7.5.0]
python-bits: 64
OS: Linux
OS-release: 5.10.0-8-amd64
machine: x86_64
processor:
byteorder: little
LC_ALL: None
LANG: en_DK.UTF-8
LOCALE: ('en_DK', 'UTF-8')
libhdf5: 1.10.6
libnetcdf: 4.8.1

xarray: 0.20.1
pandas: 1.4.1
numpy: 1.21.2
scipy: 1.7.3
netCDF4: 1.5.7
pydap: None
h5netcdf: 999
h5py: 2.10.0
Nio: None
zarr: None
cftime: 1.5.1.1
nc_time_axis: None
PseudoNetCDF: None
rasterio: None
cfgrib: None
iris: None
bottleneck: 1.3.2
dask: None
distributed: None
matplotlib: 3.5.1
cartopy: None
seaborn: None
numbagg: None
fsspec: None
cupy: None
pint: None
sparse: None
setuptools: 58.0.4
pip: 21.2.4
conda: None
pytest: None
IPython: 8.1.1
sphinx: None

Start with the DataArray.drop_sel() and drop_isel() behavior shown in the minimal reproducible example, comparing it with DataArray.sel() on the stacked 'direction' MultiIndex. Reproduce the issue with the provided xarray, pandas, and NumPy setup; done means dropping direction=(0, 0) removes only that item while preserving the other entries.