Increase minimum numpy version to 1.19.0

[Joao-Dionisio]

This version is more than 5 years old. This is to fix the pipelines (see here), following #1159

[Zeroto521]

numpy 1.19.0 is still not enough. https://github.com/scipopt/PySCIPOpt/actions/runs/21167913773/job/60877139006

tests/test_statistics.py:2: in <module>
    from helpers.utils import random_mip_1
tests/helpers/utils.py:1: in <module>
    from pyscipopt import Model, quicksum, SCIP_PARAMSETTING, exp, log, sqrt, sin
/opt/hostedtoolcache/Python/3.8.18/x64/lib/python3.8/site-packages/pyscipopt/__init__.py:11: in <module>
    from pyscipopt.scip      import Model as Model
src/pyscipopt/matrix.pxi:255: in init pyscipopt.scip
    @np.vectorize(otypes=[object], signature="(m,n),(n,p)->(m,p)")
E   TypeError: __init__() missing 1 required positional argument: 'pyfunc'
__________________ ERROR collecting tests/test_statistics.py ___________________

[Zeroto521]

Here is another way to fix this.
Use np.vectorize directly.

def _core_dot(...):
    ...
    _core_dot_nd
    ...

_core_dot_nd = np.vectorize(_core_dot_2d, otypes=[object], signature="(m,n),(n,p)->(m,p)")

def _core_dot_2d(...):
    ...

[Joao-Dionisio]

Can you please do it @Zeroto521 ?

[Zeroto521]

Does Numpy need to degrade to 1.16.0? @Joao-Dionisio

[Joao-Dionisio]

@Zeroto521 sure, why not.