test: stabilize replication/rebootstrap test

The commit 89cd7f4 ("replication: do not account booting replicas
in sync quorum") allowed replicas to leave orphan mode without waiting
for everyone listed in box.cfg.replication.

As a result, bootstrap_leader test started exiting too early, because it
relied on one of the replicas staying in orphan mode until all the
others finish bootstrap. Fix it.

While we're at it, remove the test from fragile list. It hasn't flaked
since long ago, excluding the last couple of weeks, which's fixed now.

Closes tarantool/tarantool-qa#56

NO_CHANGELOG=flaky test
NO_DOC=flaky test

(cherry picked from commit 58d9312)

Author: sergepetrenko

test/replication/rebootstrap.result

@@ -40,6 +40,9 @@ test_run:cmd('switch default')
 ---
 - true
 ...
+test_run:wait_fullmesh(SERVERS)
+---
+...
 test_run:drop_cluster(SERVERS)
 ---
 ...

test/replication/rebootstrap.test.lua

@@ -18,4 +18,5 @@ test_run:cmd('switch rebootstrap1')
 box.info.status -- running
 
 test_run:cmd('switch default')
+test_run:wait_fullmesh(SERVERS)
 test_run:drop_cluster(SERVERS)

test/replication/suite.ini

@@ -99,9 +99,6 @@ fragile = {
         "gh-5426-election-on-off.test.lua": {
             "issues": [ "gh-5506" ]
         },
-        "rebootstrap.test.lua": {
-            "issues": [ "gh-5524" ]
-        },
         "qsync_with_anon.test.lua": {
             "issues": [ "gh-5582" ]
         },