Support assignment of DOK to DOK subarray

[tbenst]

Is your feature request related to a problem? Please describe.
Allow assignment of DOK subarray without requiring creation of dense array.

Describe the solution you'd like

>>> import sparse
>>> import numpy as np
>>> array = sparse.DOK([3,4,5], dtype=np.int8)
>>> subarray = sparse.DOK([4,5], dtype=np.int8)
>>> subarray[1,1] = 1
>>> array[0] = subarray
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/tyler/lib/anaconda3/envs/glia/lib/python3.8/site-packages/sparse/_dok.py", line 317, in __setitem__
    value = np.asanyarray(value)
  File "/home/tyler/lib/anaconda3/envs/glia/lib/python3.8/site-packages/numpy/core/_asarray.py", line 136, in asanyarray
    return array(a, dtype, copy=False, order=order, subok=True)
  File "/home/tyler/lib/anaconda3/envs/glia/lib/python3.8/site-packages/sparse/_sparse_array.py", line 222, in __array__
    raise RuntimeError(
RuntimeError: Cannot convert a sparse array to dense automatically. To manually densify, use the todense method.

Describe alternatives you've considered

>>> array[0] = subarray.todense()
>>>

Additional context
If subarray is huge and very sparse, todense is not feasible.