pblimit-based masking disappears on restart

[jmarvil]

There is a type of mask created by pblimit>0 that tclean writes internal to the residual image. It shows up as a black circle in the raster display (same as in the viewer). The main issue I've noticed is that this mask can disappear if one re-runs iclean to resume an image already on disk. I also noticed that the 'hover' tool and spectrum plot show valid pixel values when hovering over this black area, but this seems less concerning. Other aspects should probably be checked (e.g. statistics, deconvolution), but I suspect they are working properly as long as the mask remains inside to the residual image. Therefore, the main purpose of this issue is related to why the internal mask goes away when restarting.