Describe the bug
mx array being first operand in arithmetic op is being converted complex numbers

To Reproduce

Include code snippet

import numpy as np
import mlx.core as mx

mx.array(1) + np.array(2)
x = mx.array(1)
y = np.array(2)
x + y 
# >>> array(3+0j, dtype=complex64)
x * y
# >>> array(2+0j, dtype=complex64)
y + x
# >>> 3
type(y + x)
# >>> <class 'numpy.int64'>
type(x + y)
# >>> <class 'mlx.core.array'>

Expected behavior
Conversion to complex should not happen. Arithmetic ops should be commutative (ignoring type like ndarray or mx).

Desktop (please complete the following information):

• OS Version: [e.g. MacOS 14.1.2] MacOS 14.1.1
• Version [e.g. 0.7.0] 0.11

Additional context
Add any other context about the problem here.