Combined resampled spectra show offset in spatial direction

[stscijgbot-jp]

Issue JP-2027 was created on JIRA by James Davies [X]:

As a follow on to JP-1829, fixing that bug revealed that the output WCS frame for combined, resampled spectra is offset spatial from what it should be. It is the correct width, but with the dataset from that example, it is offset by ~10 pixels because it uses one of the previously-computed mappings from x,y to RA,Dec and doesn't compute a new one.

This is demonstrated in new test (with the relevant part commented out) in PR

#5929

This is a bug in build_interpolated_output_wcs(). That function needs a good refactor.

While this bug is seen in MIRI LRS slit data, it is very likely also in NIRSpec FS and MOS data as well, as the same function is used. It's just much more visible in MIRI as the slit is much longer than the typical MSA slitlet.