Inspect concurrent join requests handling for CaS conflicts

A follow-up for !601 (merged)

Recently we've changed .proc_raft_join request so that it uses CaS now. The drawback is the risk of CaS rejection, which may be rather frequent since concurrent join requests are still handled by the leader in many parallel fibers.

My proposal is the following:

1. Measure the number of rejects in practice. See test/manual/test_scaling.py.
2. If it's severe, we could simply supply .proc_raft_join with a mutex.

changed milestone to %23.12.0

added backlog + 1 deleted label

assigned to @e-ivkov

mentioned in merge request !633 (merged)

Историческая справка: !633 (comment 39294)

mentioned in merge request !645 (merged)

Measure the number of rejects in practice. See test/manual/test_scaling.py.

Current stats are the following:

added in-progress label and removed backlog label

closed with merge request !645 (merged)

removed in-progress label