Memory usage too high for grid usage

On the grid we predominantly have single core jobs which are limited to using 3-4GB of memory per core. At many sites these are soft limits however if a significant number of jobs exceed the limit it quickly becomes problematic and can result in all LHCb jobs being killed at a site (and if we don't stop submitting such jobs LHCb may be banned from the site entirely).

When running the first production with Run 3 Moore in LHCbDIRAC we've noticed a high fraction of killed jobs which are likely due to memory limits being exceeded at less permissive sites. We'll need to have a solution to this before we can start having Run 3 MC productions that include Moore.

For example from https://lhcb-analysis-productions.web.cern.ch/pipelines/?id=4434884&prod=EMTF_CharmProductionXSec&job=MB_DstpToD0Pip_D0ToKsPimPip_MagDown_HLT2:

The configuration function for this job can be found here and the job can be ran locally using the command listed on the web page.