Handle asarray/asanyarray cases where like is a dask.Array

[pentschev]

Follow-up case to #8054 where the input is a Dask array and like= is another Dask array but with a different chunk type.

Before:

>>> import numpy as np, cupy as cp, dask.array as da
>>> np.asarray(da.array([1, 2, 3]), like=da.from_array(cp.array(())))
dask.array<array, shape=(3,), dtype=int64, chunksize=(3,), chunktype=numpy.ndarray>

After:

>>> import numpy as np, cupy as cp, dask.array as da
>>> np.asarray(da.array([1, 2, 3]), like=da.from_array(cp.array(())))
dask.array<asarray, shape=(3,), dtype=int64, chunksize=(3,), chunktype=cupy.ndarray>

[pentschev]

Tests failing here seem to be the same as in #8127 .

[jakirkham]

cc @jsignell (in case this is of interest 🙂)

[jsignell]

This makes sense to me!

[pentschev]

Thanks @jsignell and @jakirkham !