Test a bug found by Codenomicon where the md->st pointer was not cleared
after delete_state()

The issue happens when an IKEv1 Delete is received that contains another
Notify payload afterwards on libreswan < 3.25 that contained the support
for IMPAIR_DIE_ONINFO, since that code tried to read the deleted state.
This test is added in case there is regression and use on a deleted
state pointer would happen again - even if unlikely

This is replicated with the --impair impair-ikev1-del-with-notify

east should not crash on processing the Delete + Notify with:

Program terminated with signal SIGSEGV, Segmentation fault.
#0  0x000055b615fd8966 in informational (st=<optimized out>, md=0x55b616cad768) at /source/programs/pluto/ikev1.c:923
923				if (st != NULL &&
#0  0x000055b615fd8966 in informational (st=<optimized out>, md=0x55b616cad768) at /source/programs/pluto/ikev1.c:923
#1  0x000055b615fdb450 in process_packet_tail (mdp=mdp@entry=0x7ffcaa030988) at /source/programs/pluto/ikev1.c:2212
#2  0x000055b615fdbca7 in process_v1_packet (mdp=mdp@entry=0x7ffcaa030988) at /source/programs/pluto/ikev1.c:1728