[NDArray] Add __setitem__ overload for scalar

[shivasankarka]

This PR implements the following methods to support arr[0:1, 3:4] = 10.0,

def __setitem__(
        mut self, *slices: Variant[Slice, Int], scalar: Scalar[Self.dtype]
    ) raises:

def __setitem__(
        mut self, *slices: Slice, scalar: Scalar[Self.dtype]
    ) raises:

def _setitem_slice_scalar(
        mut self, slices: List[Slice], val: Scalar[Self.dtype]
    ) raises

This let's the user do,

def main() raises:
    var arr = nm.linspace[nm.f32](0, 9, 10).reshape(Shape(2, 5))
    print(arr)

    arr.__setitem__(0, Slice(0, 3, 1), scalar=Scalar[nm.f32](42.0))
    print(arr)

Note:

• Unfortunately overload resolution in Mojo isn't fixed yet. So we have to currently call __setitem__ explicitly with scalar keyword :(. Hopefully once that's fixed, we should be able to do arr[0:2, 2] = 10.0.

[forfudan]

I think maybe we can even create a normal function without dunders if overload is not possible in Mojo.

[shivasankarka]

@forfudan I was thinking the same. Should we go with .get and .set methods? In which case, user can at least do something like

arr.get(Slice(0, 2), 1) ->  Get an array
arr.set(Slice(0, 2), 1, val=10.0) -> Set scalar onto the indexed coordinates
arr.set(Slice(0, 2), 1, arr=array) -> Set an array onto the indexed coordinates