dp3_max_threads not passed to certain DP3 calls in HBA target
I noticed certain DP3 steps like e.g. dp3_make_parset_target.cwl don't use the numthreads passed to them and still tried to use all 96 CPUs on my node at times. It could run many chunks in parallel, which caused high (>500% in this case) system load. I can make a MR addressing the places I noticed it happening.