Regression in pixel to world mapping for lookup tables

[nmearl]

Looks like this change introduced an issue when using Tabular1D lookup tables:

Testing the code block below

def gwcs_from_array(array):
    coord_frame = cf.CoordinateFrame(naxes=1,
                                     axes_type=('SPECTRAL',),
                                     axes_order=(0,))
    spec_frame = cf.SpectralFrame(unit=array.unit, axes_order=(0,))

    forward_transform = Tabular1D(np.arange(len(array)),
                                          lookup_table=array)
    forward_transform.inverse = Tabular1D(
        array, lookup_table=np.arange(len(array)))

    tabular_gwcs = GWCS(forward_transform=forward_transform,
                                input_frame=coord_frame,
                                output_frame=spec_frame)

    return tabular_gwcs

for gwcs 0.15 yields

In [3]: wcs = gwcs_from_array(np.arange(100) * u.AA)

In [4]: wcs.pixel_to_world([10, 11])
Out[4]: <Quantity [10., 11.] Angstrom>

and for gwcs 0.16 yields

In [3]: wcs = gwcs_from_array(np.arange(100) * u.AA)

In [4]: wcs.pixel_to_world([10, 11])
Out[4]: <Quantity 10. Angstrom>

[nden]

Change that caused regression was reverted and a new patch release is out - 0.16.1