Turn numpy warnings into numpy errors?

[hsl-petrhaj]

@johpiip @zptro
Do you think it makes sense to introduce the following line that promotes numpy errors from warnings to errors?
I think it is very strange that the program continues even after dividing by zero for example.

This continues like nothing has happenned:

import numpy as np
a = np.array([1,2,3]) 
b = np.array([0,0,0])
c = a/b

This throws an error:

import numpy as np
np.seterr(all='raise')

a = np.array([1,2,3]) 
b = np.array([0,0,0])
c = a/b

Of course that would push us to pay more attention to numpy warnings which could also be somewhat annoying. But, safety first?

[johpiip]

Sounds like a good idea to me! You should test it first with our official scenarios that they do not raise errors - and if they do, we should think how to fix them first.