lazy.utils.resample output dtype

[wyli]

Describe the bug
the output of grid resampling is always float32

MONAI/monai/transforms/lazy/utils.py

Lines 226 to 227 in 6a7f35b

	with resampler.trace_transform(False): # don't track this transform in `img`
	return resampler(img=img, **call_kwargs)

MONAI/monai/transforms/spatial/functional.py

Line 182 in 6a7f35b

out = _maybe_new_metatensor(img, dtype=torch.float32)

the crop pad 'resampling' should also output float32

MONAI/monai/transforms/lazy/utils.py

Lines 220 to 222 in 6a7f35b

	img = monai.transforms.crop_or_pad_nd(img, matrix_np, out_spatial_size, mode=call_kwargs["padding_mode"])
	img.affine = call_kwargs["dst_affine"]
	return img

otherwise collating random transforms of different data types can run into exceptions.

cc @dongyang0122

[wyli]

this is still an issue for the early returning cases

MONAI/monai/transforms/lazy/utils.py

Lines 218 to 219 in 57c618c

	img.affine = call_kwargs["dst_affine"]
	return img