Broadcast does not return Datasets with unified chunks

What happened:
If I broadcast a Dataset with chunked DataArrays, the resulting DataArrays are chunked differently.

What you expected to happen:
If I broadcast a dataset with 2 vectors of chunk size 1, I'd expect to get 2D arrays with chunksize (1, 1), rather than (1, N) and (M, 1).

Minimal Complete Verifiable Example:

import xarray as xr
ds = xr.Dataset({i: xr.DataArray(range(i*10), dims=f"dim_{i}").chunk(1) for i in range(1, 3)})
ds.broadcast_like(ds)

<xarray.Dataset>
Dimensions:  (dim_1: 10, dim_2: 20)
Dimensions without coordinates: dim_1, dim_2
Data variables:
    1        (dim_1, dim_2) int64 dask.array<chunksize=(1, 20), meta=np.ndarray>
    2        (dim_1, dim_2) int64 dask.array<chunksize=(10, 1), meta=np.ndarray>

Anything else we need to know?:

Environment:

Start by running the minimal example through Dataset.broadcast_like and inspect the resulting DataArray chunks. The fix is complete when broadcasting the two chunked vectors produces 2D arrays with chunksize (1, 1), as expected in the issue.