multiple pdu sessions regression fix

Since E1AP MR merge, the 2 pdu sessions required by several commercial UEs was not working anymore
the issue is not very clear, anyway the e1 code has partially hardcoded one single pdu session
This MR fixes the "monolithic" case and likely also the CU/DU but not the dis-aggregated gNB with also E1

I will make more MRs to simplify the code, remove most of global variables that are very confusing for any code modification