Round tripping Zarr datasets with scaling

What happened?

Data was loaded into Xarray via OpenDataCube dc.load method. This Xarray was persisted to Zarr using

xx.time.encoding['units'] = "seconds since 1970-01-01 00:00:00"
xx.time.attrs = {}
xx.to_zarr('test.zarr',mode='w',consolidated=True)

When this zarr is loaded back using

ds_z = xr.open_dataset('test.zarr',  engine = 'zarr', consolidated = True)

The round tripped dataset has scaling applied with encodings saved in bands e.g. ds_z.red.encoding as

ds_z.red.encoding
{'chunks': (1, 610, 522),
 'preferred_chunks': {'time': 1, 'y': 610, 'x': 522},
 'compressor': Blosc(cname='lz4', clevel=5, shuffle=SHUFFLE, blocksize=0),
 'filters': None,
 'scale_factor': 0.0001,
 'add_offset': -0.1,
 'dtype': dtype('uint16'),
 'coordinates': 'spatial_ref'}

The values returned do not have the scaling auto-applied.

What did you expect to happen?

Scaling to DataArrays is auto-applied when rounding tripping to-from Zarr via Xarray.

Minimal Complete Verifiable Example

ds.time.encoding['units'] = "seconds since 1970-01-01 00:00:00"
ds.time.attrs = {}
ds.to_zarr('test.zarr',mode='w',consolidated=True)

ds_z = xr.open_dataset('test.zarr',  engine = 'zarr', consolidated = True)

MVCE confirmation

• Minimal example — the example is as focused as reasonably possible to demonstrate the underlying issue in xarray.
• Complete example — the example is self-contained, including all data and the text of any traceback.
• Verifiable example — the example copy & pastes into an IPython prompt or Binder notebook, returning the result.
• New issue — a search of GitHub Issues suggests this is not a duplicate.
• Recent environment — the issue occurs with the latest version of xarray and its dependencies.

Relevant log output

No response

Anything else we need to know?

No response

Environment

Start by reproducing the minimal example using to_zarr and xr.open_dataset with the supplied environment details, then inspect how the scale_factor and add_offset encodings are written and read. Confirm the expected decoded values after round-tripping and add coverage for the demonstrated scaling case; done means the reopened DataArray automatically applies the stored scaling.