Fix false rejection of rms map

[AlexKurek]

check_rmsmap evaluates standard deviation (stdsub) on the interpolated pixel array (rms). Interpolation acts as a low-pass filter, depressing pixel variance below the expectation (rms_expect). Consequently, stdsub frequently falls below the 1.1 * rms_expect threshold, causing varying RMS maps to be rejected and replaced with a constant (use_rms_map = False).

Fix
Subsampled the rms map and mask_arr at step increments corresponding to the evaluation grid step size (img.rms_box[1]) before computing stdsub.

[gmloose]

I don't understand your reasoning. Why does sub-sampling avoid interpolation? Why would calculating the standard deviation on the original data cause low-pass filtering?

[AlexKurek]

rms array passed to check_rmsmap is not the original data. It has been interpolated to the full image resolution via mapcoord_threaded (3rd-order spline filter). Interpolation suppresses variance.

Because the variance is lowered, stdsub drops frequently below the 1.1 * rms_expect threshold. This causes varying RMS maps to be rejected.

By sub-sampling the map at the original rms_box[1] step size, we are evaluating the standard deviation primarily on the nodes of the original coarse grid. This restores the true variance for the threshold check.