Exception after DAG optimization

[tomwhite]

>>> import cubed.array_api as xp
>>> a = xp.ones((2, 2), chunks=(2, 2))
>>> b = xp.positive(a)
>>> c = xp.asarray(b)
>>> d = xp.equal(b, c)
>>> d.compute()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/tom/projects-workspace/barry/cubed/core/array.py", line 77, in compute
    self.plan.execute(
  File "/Users/tom/projects-workspace/barry/cubed/core/plan.py", line 188, in execute
    executor.execute_plan(plan, **kwargs)
  File "/Users/tom/opt/miniconda3/envs/barry/lib/python3.8/site-packages/rechunker/executors/python.py", line 29, in execute_plan
    plan()
  File "/Users/tom/opt/miniconda3/envs/barry/lib/python3.8/site-packages/rechunker/executors/python.py", line 22, in plan
    stage.function(m, config=pipeline.config)
  File "/Users/tom/projects-workspace/barry/cubed/primitive/blockwise.py", line 36, in apply_blockwise
    config.write.array[out_chunk_key] = config.function(*args)
  File "/Users/tom/projects-workspace/barry/cubed/primitive/blockwise.py", line 230, in fused_func
    return pipeline2.config.function(pipeline1.config.function(*args))
TypeError: equal() takes from 2 to 3 positional arguments but 1 were given
>>> d.compute(optimize_graph=False)
array([[ True,  True],
       [ True,  True]])

What is happening is that asarray returns b, and then b and d are fused, which causes a problem with the number of arguments.

This is similar to test_no_fusion. Found as array_api_tests/test_creation_functions.py::test_asarray_arrays failure.