Scalars in shape definition

[TizianoDeMatteis]

Describe the bug

If a container shape is defined by using scalars, the produced SDFG will not generate proper code

To Reproduce

Consider the following example:

import dace
import numpy as np

A, B, C, D, E = (dace.symbol(s, dace.int32) for s in ('A', 'B', 'C', 'D', 'E'))


@dace.program
def init_array():
    output_height = ((A - C) + 1)
    output_width = ((B - D) + 1)
    # This does not work
    container = np.zeros((A * B, output_height * output_width * E))
    # This works
    # container = np.zeros((A * B, ((A - C) + 1) * ((B - D) + 1) * E))
    return container


A = 32
B = 32
C = 4
D = 4
E = 2
sdfg = init_array.to_sdfg()
container = sdfg(A=A, B=B, C=C, D=D, E=E)

In the generated SDFG, the variable __tmp3 is not defined anywhere.
The definition that just uses symbols works.

Expected behavior
Functional generated code.

[tbennun]

This should be fixed the same way other frontends handle this: Start a new state, add __sym_x = x interstate assignments, followed by using __sym_x for the sizes. I can confirm that simplify successfully cleans those situations up if applicable. This should probably be something more substantial/refactored in the frontend that we can call whenever we define a new data container in replacements.