Small fixes on image.extract_patches

[guillaumebaquiast]

Two small fixes on extract_patches:

• Passing argument size as an int currently fails for symbolic tensor:

>>> x = KerasTensor([None, 20, 20, 3])
>>> keras_core.ops.image.extract_patches(x, size=5)
Traceback (most recent call last):
...
    strides = (self.size[0], self.size[1])
TypeError: 'int' object is not subscriptable

Fixed and added unit-test.

• Examples in docstring are failing for me, I have to convert the arrays to "float32" beforehand.

>>> image = np.random.random((1, 20, 20, 3)) # batch of 2 RGB images
>>> patches = keras_core.ops.image.extract_patches(image, (5, 5))
Traceback (most recent call last):
...
tensorflow.python.framework.errors_impl.InvalidArgumentError: cannot compute Conv2D as input #1(zero-based) was expected to be a double tensor but is a float tensor [Op:Conv2D] name:

I edited the docstring accordingly.

[fchollet]

LGTM, thanks for the fix!