Numpy fill accepts only numbers

[TizianoDeMatteis]

Describe the bug

Current replacement for numpy.fill checks that the value is a Number (or boolean). This prevents using a variable.

To Reproduce

import dace
import numpy as np

N = dace.symbol("N", dace.int32)


@dace.program
def fill_dace(A: dace.float64[N, N], value: dace.float64):
    A.fill(value)


N = 8
value = 0.3
input = np.ndarray((N, N)).astype(np.float64)
input_ref = np.copy(input)

fill_dace(input, value)
input_ref.fill(value)
assert np.allclose(input, input_ref)

Steps to reproduce the behavior:

1. Run the code above
2. It will return a DaceSyntaxError

Expected behavior
Allow also variables (or parameters) to be passed to fill

[philip-paul-mueller]

I will take a look at it.

[philip-paul-mueller]

The reason for this behaviour is located in _elementwise() (to which fill() is forwarded).
It directly creates a very simple map with a single tasklet.
I think that this rerstriction is because it could lead to arbitrary complex map internals otherwise.

I think that a "simple" maybe not good way would be to rewrite it to an assignement.