Array API: Combiner.sigma_clipping and the default sigma_func densify via astropy.stats

### Problem

Two closely related densification points in the combining pipeline:

1. `Combiner.sigma_clipping` delegates to `astropy.stats.sigma_clip`, which internally does `np.asanyarray` and works with `np.ma` masked arrays — both densify non-numpy input:

https://github.com/astropy/ccdproc/blob/9d25eeefda8a38fc442c1a18e79ca2fa8ca40559/ccdproc/combiner.py#L449

2. `ccdproc.sigma_func` — the **default** deviation function for `median_combine` uncertainty — is built on `astropy.stats.median_absolute_deviation`, which is likewise numpy-only:

https://github.com/astropy/ccdproc/blob/9d25eeefda8a38fc442c1a18e79ca2fa8ca40559/ccdproc/core.py#L1233

These likely account for a substantial share of the `Implicit conversion to a NumPy array is not allowed` failures in #911.

### Options

- Provide xp-native implementations in ccdproc: an iterative sigma-clip using `xp.where`, and a MAD built on the fallback median discussed in #906.
- Or wait for / contribute array-API support in `astropy.stats` (tracked in #940).

---
Found during a review of the array API implementation from #885; follow-up to #909 / #911. Related: #906.

Start in ccdproc/combiner.py at Combiner.sigma_clipping and ccdproc/core.py at sigma_func, then review the linked astropy.stats behavior and related issues #906, #911, and #940. Determine whether both densification points should gain array-API support in ccdproc or wait for astropy; done requires an agreed implementation path that avoids the reported NumPy conversion failures.

Written by the indexing model from the issue text.