concatenate() should handle scalar symbols

[arcondello]

NumPy's concatenate() does not handle scalars

In [2]: np.concatenate([0, 1])
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[2], line 1
----> 1 np.concatenate([0, 1])

File <__array_function__ internals>:180, in concatenate(*args, **kwargs)

ValueError: zero-dimensional arrays cannot be concatenated

I assume this is because you can simply do

In [11]: np.asarray([0, 1])
Out[11]: array([0, 1])

However, for our purposes it would be convenient to handle it. So we can do constructions like

import numpy as np

from dwave.optimization import concatenate, Model

model = Model()

symbols = [model.constant(0), model.constant(1)]

concatenate(symbols)  # this doesn't work

for now it can be done with

import numpy as np

from dwave.optimization import concatenate, Model

model = Model()

symbols = [model.constant(0), model.constant(1)]

concatenate(tuple(s.reshape((1,)) for s in symbols))  # this works

[arcondello]

I should be less cryptic. Thinking about it more, I think what I actually want is numpy.stack(). That is, I want to combine $n$ 0d arrays (i.e. scalars) into a length $n$ 1d array. This can either be handled at the Python level via reshapes and concatenates or at the C++ level.

[samuelbodin]

Implementing stack would be the better approach imo. But maybe a quicker solution for now is to go with the Python. Similar to what is in my PR, only that we should explicitly check that ndim is 0 before reshaping and not limit it to constants since we'd also probably want concatenate([model.binary(), model.binary()]) to work