Expand types of `Dataset` keys that can be saved to disk

Is your feature request related to a problem?

For a few applications I can think of (e.g. OME-Zarr, single cell data) I (and others) have commonly used integers as the key.

Currently doing this:

import xarray as xr

xr.Dataset({1: xr.DataArray([1, 2])}).to_zarr("test.nc")

Gives this error:

TypeError: Invalid name 1 for DataArray or Dataset key: must be either a string or None for serialization to netCDF or zarr files

Describe the solution you'd like

Xarray handles this behind the scenes for me, saving that key as a str and encoding in the zarr metadata that on open_dataset it should be restored to an int. This could be generalized to other types as well.

Describe alternatives you've considered

Making an accessor that handles that conversion, but I'd prefer it just work for as many cases as practicable.

Additional context

No response

Review the Dataset serialization paths used by to_zarr and open_dataset, beginning with the key validation that raises the shown TypeError. Define supported key types and verify round-trip restoration through zarr metadata, including the integer-key example; the issue names no specific files or tests.