[Jepsen] [YSQL] Postgres processes leftover when using LXC nodes

[frozenspider]

Jira Link: DB-2487
This is very similar to #2075, but is only reproducible when using LXC nodes instead of real nodes.
Using the same steps as described in #2075 (except not using real nodes), logging number of postgres processes via ps aux | grep postgres | wc -l hourly using cron job, and killing all with sudo pkill -9 postgres before starting Jepsen run, I observed number of postgres processes on a host node rising from 30 to 800, with number of per-LXC postgres processes increasing from 7 to 180.
Most of those processes were stuck in authentication state, and some - in SELECT or INSERT. This is also different to the actual issue reported in #2075