f32 is not infered on basic mult

Consider this snippet:
```
import numpy as np
import dace

@dace.program
def dace_float_cast(A, B, timestep: dace.float32):
dt2: dace.float32 = 2 * timestep
A[:] = B[:] * dt2

A = np.zeros(10, dtype=np.float32)
B = np.zeros(10, dtype=np.float32)
t = np.float32(250.0)

dace_float_cast(A, B, t)
```

SDFG & code generated (below) show `dt2` is consider a 64-bit float, despite the cast.

```
...
double dt2;
{
float __in2 = timestep;
double __out;

///////////////////
// Tasklet code (_Mult_)
__out = (dace::float64(2) * dace::float64(__in2));
///////////////////

dt2 = __out;
}
...
```

SDFG (not a real zip, just going around GitHub) [program.sdfg.zip](https://github.com/spcl/dace/files/13814316/program.sdfg.zip)

Start by running the provided dace_float_cast example and inspecting program.sdfg.zip alongside the generated code. Trace how the annotated timestep and the multiplication by 2 are typed, then verify that the generated dt2 and its arithmetic use float32 rather than float64.

Written by the indexing model from the issue text.