UPG crash on multi-core setup in proxy mode

[sergeymatov]

Under high traffic rate and session churn loop UPG crashed

#0  0x00007fcfdd50c18b in raise () from /lib/x86_64-linux-gnu/libc.so.6
#1  0x00007fcfdd4eb859 in abort () from /lib/x86_64-linux-gnu/libc.so.6
#2  0x000055e2b7a7c905 in os_exit (code=<optimized out>) at /src/vpp/src/vpp/vnet/main.c:433
#3  0x00007fcfdddd0d18 in unix_signal_handler (signum=<optimized out>, si=<optimized out>, uc=<optimized out> at /src/vpp/src/vlib/unix/main.c:184
#4  <signal handler called>
#5  proxy_start_connect_fn (session_index=0x1300e4fe9) at /src/vpp/src/plugins/upf/upf_proxy.c:383
#6  0x00007fcfdf7fd6b4 in vl_api_rpc_call_t_handler (mp=0x1300e4fd0) at /src/vpp/src/vlibmemory/vlib_api.c:531
#7  0x00007fcfdf806b61 in vl_msg_api_handler_with_vm_node (am=0x7fcfdf814c90 <api_global_main>, vlib_rp=<optimized out>, the_msg=0x1300e4fd0, vm=<optimized out>,
node=0x7fce5dbb1ec0, is_private=0 '\000') at /src/vpp/src/vlibapi/api_shared.c:635
#8  0x00007fcfdf7e8c48 in vl_mem_api_handle_rpc (vm=0x7fcfddded800 <vlib_global_main>, node=0x7fce5dbb1ec0) at /src/vpp/src/vlibmemory/memory_api.c:746
#9  0x00007fcfdf7f763b in vl_api_clnt_process (vm=0x7fcfddded800 <vlib_global_main>, node=0x7fce5dbb1ec0, f=<optimized out>) at /src/vpp/src/vlibmemory/vlib_api.c:337`

It seems dedicated upf session has been flushed, but traffic is still hitting upg proxy callbacks in separate thread