Update numpy, pint, python, scipy, versioneer

[thehrh]

All of the remaining dependencies are brought up to date (also see the linked issues). If I am not mistaken, none of these changes are expected to break existing PISA installations, and indeed don't seem to be doing so (as confirmed by successful unit and service tests in environments of mine without updated dependencies).

Note: Select "squash and merge" if and when this is merged to create a single commit.

• resolves Update numpy from <1.23 #900
• resolves Update pint from <= 0.19 #901
• resolves Update scipy from < 1.14 #902
• resolves Update python from <= 3.10 #904
• resolves Update versioneer from release 0.16 to 0.29 #907

[thehrh]

I doubt this ever worked the way it was intended to, i.e., managed to only detect an ndarray that was created with uncertainties.

The reason is that np.issubsctype(obj, AffineScalarFunc) would determine the dtype of obj, but because AffineScalarFunc isn't a NumPy scalar subclass, the ndarray obj would have dtype=np.object_. Hence, the check couldn't detect whether the elements are AffineScalarFunc instances (and not, e.g., PISA Params, and would always return True. As shown below, np.issubsctype(np.object_, AffineScalarFunc) is True, as is np.issubdtype(np.object_, AffineScalarFunc)---probably because np.obj2sctype(AffineScalarFunc) yields np.object_.

Applying this check to an array with uncertainties created in two different ways:

>>> import numpy as np
>>> from pisa.core.param import Param
>>> from uncertainties import ufloat, unumpy
>>> from uncertainties.core import AffineScalarFunc
>>> arr1 = unumpy.uarray([1, 2], [0.01, 0.002])
>>> print(arr1.dtype)
object
>>> isinstance(arr1, np.ndarray)
True
>>> arr2 = np.array([ufloat(1, 0.1), ufloat(2, 0.002)])
>>> print(arr2.dtype)
object
>>> isinstance(arr2, np.ndarray)
True
>>> u1 = ufloat(1, 0.1)
>>> isinstance(u1, AffineScalarFunc)
True
>>> np.issubsctype(arr1, AffineScalarFunc)
True
>>> np.issubsctype(arr2, AffineScalarFunc)
True
>>> # but also:
>>> p = Param(name="x", value=0.1, prior=None, range=[0,1], is_fixed=True)
>>> arr3 = np.array([p, p])
>>> np.issubsctype(arr3, AffineScalarFunc)
True

[thehrh]

On a related note, in old numpy (in my case 1.22.4), np.issubdtype(arr1.dtype, AffineScalarFunc) still worked, while now in numpy 2.3.5 it doesn't any longer (raises ValueError: dtype attribute is not a valid dtype instance for the second argument, AffineScalarFunc).