Fix transposed axes in make_image_interpolations

[astronomyk]

Yet again, just a bit of clean up of a latent bug that has not yet bitten.

TL;DR - aligns the RectBivariateSpline shape to the axis convention of .rectify

NAXIS1, NAXIS2 --> NAXIS2, NAXIS1 in spectral_trace_list_utils.py

From the below mentioned testing notebook:

from astropy.io import fits
from scipy.interpolate import RectBivariateSpline

# a non-square "detector window", e.g. a readout sub-window
xx, yy = np.meshgrid(np.arange(300), np.arange(200))
img = np.sin(xx / 20) * np.cos(yy / 15)          # shape (200, 300)
hdu = fits.ImageHDU(data=img)

# --- removed construction ---
try:
    RectBivariateSpline(np.arange(hdu.header["NAXIS1"]),
                        np.arange(hdu.header["NAXIS2"]),
                        hdu.data, kx=1, ky=1)
except ValueError as err:
    print("old construction raises:", err)

old construction raises: x dimension of z must have same number of elements as x

---

What

make_image_interpolations constructed its splines as RectBivariateSpline(arange(NAXIS1), arange(NAXIS2), hdu.data) — axes transposed relative to the data shape (NAXIS2, NAXIS1). For any non-square image this raises

ValueError: x dimension of z must have same number of elements as x

It has gone unnoticed because the detector images we currently rectify are square (2048x2048), and because the caller in SpectralTrace.rectify compensates by evaluating with swapped arguments, interps[ihdu](jarr, iarr).

The spline is now built with axes matching the data layout, so the (j, i) call convention used by rectify (and by the existing accuracy test) is correct by construction; the convention is documented in the docstring. Results on square images are bit-identical.

Testing

• New regression test TestImageInterpolations::test_works_with_non_square_image (80x50 image), mirroring the existing accuracy test.
• A notebook reproducing the old failure and verifying bit-identical results on square images is attached below.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.50%. Comparing base (c47dc7a) to head (922a929).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #984   +/-   ##
=======================================
  Coverage   76.50%   76.50%           
=======================================
  Files          69       69           
  Lines        9062     9062           
=======================================
  Hits         6933     6933           
  Misses       2129     2129           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[oczoske]

This is okay. The same issue has been noticed and fixed by a human in #950 for PSF interpolation.