make sure _onenorm_matrix_power_nnm actually returns a float

[nschloe]

The function _onenorm_matrix_power_nnm() computes the 1-norm of a non-negative integer power of a non-negative matrix and

Returns
    -------
    out : float
        The 1-norm of the matrix power p of A.

However, it really does return numpy.array([alpha]) with alpha being the 1-norm. This PR fixes things by actually returning a float.

Before:

print(_onenorm_matrix_power_nnm(A, p))   # array([1.31])

After:

print(_onenorm_matrix_power_nnm(A, p))   # 1.31

The change is probably too small to warrant an explicit test.

[rgommers]

The change is probably too small to warrant an explicit test.

It would be good to at least show the old and new behavior in the PR description. From a quick look it looks like this affects the behavior of expm. Not sure about the other variants of expm?

[nschloe]

@rgommers I've extended the description to highlight the difference.

[pv]

IIUC, in practice this change has no effect, but np.max was probably intended here.

[rgommers]

@nschloe are you using that private function directly? The only public one that uses it seems to be expm, however that returns a matrix:

In [19]: A
Out[19]: 
<1x1 sparse matrix of type '<class 'numpy.int64'>'
	with 1 stored elements in Compressed Sparse Column format>

In [20]: A.todense()
Out[20]: matrix([[1]], dtype=int64)

In [21]: sl.expm(A).todense()
Out[21]: matrix([[2.71828183]])

[nschloe]

@rgommers Not really, it's just something I noticed when I fiddled with numpy for other reasons. It seemed easy enough to fix it, so that's what I did.

[rgommers]

Okay, thanks for clarifying. Since it doesn't change any user-visible behavior and np.max was intended, let's just merge this.

Thanks @nschloe, @pv