Staging instance, all changes can be removed at any time

[dynamic Infra] Add an upper memory limit on loader deployments

In order to avoid some loader memory bomb and secure the cluster, we should add an upper memory limit to avoid some pods to consume too much memory.

The first candidates could be the save code now and the add forge now deployment. The svn loaders look also big consumers.

memory

The nodes are prone to crash when their 256Go of memory are fully consumed.