Array API: wcs_project mixes numpy output from reproject with xp operations

[mwcraig]

Problem

In wcs_project, reproject_interp returns numpy arrays regardless of the input library, but the subsequent mask handling uses the input's namespace:

ccdproc/ccdproc/core.py

Lines 1163 to 1180 in 9d25eee

	projected_image_raw, _ = reproject_interp(
	(ccd.data, ccd.wcs), target_wcs, shape_out=target_shape, order=order
	)
	reprojected_mask = None
	if ccd.mask is not None:
	reprojected_mask, _ = reproject_interp(
	(ccd.mask, ccd.wcs), target_wcs, shape_out=target_shape, order=order
	)
	# Make the mask 1 if the reprojected mask pixel value is non-zero.
	# A small threshold is included to allow for some rounding in
	# reproject_interp.
	reprojected_mask = reprojected_mask > 1e-8
	# The reprojection will contain nan for any pixels for which the source
	# was outside the original image. Those should be masked also.
	output_mask = xp.isnan(projected_image_raw)

With xp = cupy, xp.isnan(projected_image_raw) receives a numpy array — CuPy functions reject numpy input (TypeError: 'a' must be a cupy.ndarray). Even where it happens to work, the returned CCDData holds numpy data while the caller supplied another library's array, silently switching namespaces mid-pipeline. Likely contributes to the WCS-related failures in #910.

Suggested direction

reproject is CPU-only, so this falls under the general policy decision for CPU-only operations (#935): either convert explicitly to host, reproject, and convert the result (data and mask) back to the input namespace — or document wcs_project as numpy-only and raise an informative error.

---

Found during a review of the array API implementation from #885; follow-up to #909 / #910.