AST-1558: add fourth selfcal loop

Add fourth loop including calibrate_4, image_4 and optionally predict_4. In a new Rapthor update, the predict step is not happening anymore after the first selfcal iteration, hence I added the configurability to skip predict from the third cycle onwards via the 'subtract_bright_sources' parameter.

I did run a complete pipeline on midbands_averaged.ms and midbands.ms, and the RMS value on the image of the 4th selfcal cycle is higher than the 3rd one. I also noticed that the same happened in the original Rapthor run, so I will not consider it as a problem. The amount of sources being detected after the 4 cycle increases so that is a sign that more and more sources are being detected.