[BUG] `mx.load` misinterprets dtype as `complex64` instead of `float64`

[krishi-saripalli]

Describe the bug
mx.load() appears to misinterpret the datatype of .npy files with 64-bit data as complex64 when it should be float64.

To Reproduce

Include code snippet

>>> import numpy as np
>>> import mlx.core as mx
>>> data = np.load("my_data.npy")
>>> print(f"Dtype: {data.dtype}")
Dtype: float64
>>> mx_data = mx.load("my_data.npy")
>>> print(f"Dtype: {mx_data.dtype}")
Dtype: mlx.core.complex64
>>> 

Expected behavior
Like numpy, the data should be correctly loaded as float64

Desktop (please complete the following information):

• OS Version: MacOS 15.6
• Version 0.30.1

Additional context
Add any other context about the problem here.

[MillaFleurs]

Can you provide a copy of my_data.npy to make it easier to reproduce?

[awni]

This is already fixed in #2960