Boolean index assignment fails for values of `ndim>1`

This works in NumPy, but fails in Dask.

```python
In [1]: import numpy as np

In [2]: import dask.array as da

In [3]: x = np.asarray([[1, 2]])

In [4]: y = np.asarray([[3, 4]])

In [5]: i = np.asarray([True])

In [6]: xd = da.asarray(x)

In [7]: yd = da.asarray(y)

In [8]: id = da.asarray(i)

In [9]: x[i] = y[i]

In [10]: x
Out[10]: array([[3, 4]])

In [11]: xd[id] = yd[id]
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
Cell In[11], line 1
----> 1 xd[id] = yd[id]

File ~/programming/pixi-dev-scipystack/scipy/.pixi/envs/array-api/lib/python3.12/site-packages/dask/array/core.py:1903, in Array.__setitem__(self, key, value)
1900 from dask.array.routines import where
1902 if isinstance(value, Array) and value.ndim > 1:
-> 1903 raise ValueError("boolean index array should have 1 dimension")
1904 try:
1905 y = where(key, value, self)

ValueError: boolean index array should have 1 dimension

In [12]: yd[id].ndim
Out[12]: 2

```

**Anything else we need to know?**:
I think the error message is unclear too. I initially thought that the "boolean index array" referred to `id` here (the index array), as opposed to `yd[id]`, the value array being assigned.

Start at dask/array/core.py around the __setitem__ path shown in the traceback, then reproduce the xd[id] = yd[id] example from the issue. Compare the assignment with NumPy's result and clarify the reported error; done means the ndim>1 value assignment succeeds consistently with NumPy and the error wording no longer misidentifies the boolean index array.

Written by the indexing model from the issue text.