Flipped columns/rows in overscan correction

[ben-cassese]

In wcs.locate.get_observability_mask, the criteria for checking whether a coordinate is visible or not is given by:

C, R = wcs._all_world2pix(
            np.asarray([coords.ra.deg[k], coords.dec.deg[k]]).T,
            0,
            tolerance=0.0001,
            maxiter=20,
            adaptive=False,
            detect_divergence=False,
            quiet=True,
        ).T
# Update the mask with those pixels
j = (
    (C >= 0)
    & (R >= 44)
    & (C <= (wcs._naxis[0] - 30))
    & (R <= (wcs._naxis[1] - 44))
)

where j is the flag for observable or not. However, the 2-sided 44 pixel overscan should be applied to the columns, not rows, while the 30 pixel buffer should apply to the rows on the top of the detector, not the columns. Picking a random point in that's visually within the overscan region of an FFI from Sector 1, Camera 4, CCD 4, (RA=97.7, Dec=-62.62), this function incorrectly flags that it would be observable. Flipping the logic to:

j = (
    (C >= 44)
    & (R >= 0)
    & (C <= (wcs._naxis[0] - 44))
    & (R <= (wcs._naxis[1] - 30))
)

correctly flags that it's not visible.