Annulus stats with robust sigma clipping, suitable e.g. for better local background estimation in crowded cases:

• Added sep.stats_circann to compute annulus statistics (mean, std, median, mad_std, sigma‑clipped mean, and area) with the same mask/segmap semantics as existing aperture routines.
• Sigma‑clipping uses median/MAD (scaled by 1.4826, so like mad_std) and exposes clip_sigma/clip_iters for configuration.

Also, a couple of minor fixes:

• Avoid divide‑by‑zero when subpix == 0 by only computing scale/offset when subpix > 0 (src/aperture.i).
• When mask covers the full aperture, return zeros and set SEP_APER_ALLMASKED instead of dividing by zero (src/aperture.i).
• sum_circann docstring now matches the signature order (var, err) (sep.pyx).