test: replication/transaction.test.lua flaky fails in parallel

[avtikhon]

Tarantool version:
master

OS version:
All

Bug description:
Error:

[015] Test failed! Result content mismatch:
[015] --- replication/transaction.result	Wed Jun 26 10:31:01 2019
[015] +++ replication/transaction.reject	Wed Jun 26 10:52:26 2019
[015] @@ -94,7 +94,7 @@
[015]  -- replication stopped of third transaction
[015]  v1[1] + 2 == box.info.vclock[1]
[015]  ---
[015] -- true
[015] +- false
[015]  ...
[015]  box.space.test:select()
[015]  ---
[015] 

Appeared after the commit:

commit 8c84932ade1b70971890dfd27d72b3145834594b (origin/master, master)
Author: Georgy Kirichenko <georgy@tarantool.org>
Date:   Fri Jun 7 22:24:20 2019 +0300

    applier: apply transaction in parallel
    
    Applier use asynchronous transaction to batch journal writes. All
    appliers share the replicaset.applier.tx_vclock which means the vclock
    applied but not necessarily written to a journal. Appliers use a trigger
    to coordinate in case of failure - when a transaction is going to
    be rolled back.
    
    Closes: tarantool/tarantool#1254

Steps to reproduce:
./test-run.py -j 15 for r in {1..15} ; do echo replication/transaction.test.lua ; done 2>/dev/null

Optional (but very desirable):

• coredump
• backtrace
• netstat

[avtikhon]

Still fails on 2.4.0-16-gcdf502c66 in %2 of runs of:

while ./test-run.py -j 30 `for r in {1..100} ; do echo replication/transaction.test.lua ; done` 2>/dev/null ; do n=$(($n+1)) ; echo RUN ========== $n ; done