Applier remains stopped after replica rebootstrap

After an unrecoverable replication error, such as ER_TUPLE_FOUND or ER_SPLIT_BRAIN, the applier is stopped. If the user rebootstraps the failed replica by wiping its data files and restarting it with the same UUID, the rebootstrap succeeds, but the applier on the other instance remains stopped.

As a result, replication is restored only in one direction. The user has to restart replication manually, which is not obvious: the same error remains visible after rebootstrap, so the user may assume that the rebootstrap failed and try restarting or rebootstrapping the replica again.

We should probably add a mechanism for automatically restarting the applier in this case.

Reproducer:

-- Instance 1.
-- Step 1.
box.cfg{
    listen = '127.0.0.1:3301',
    replication = {
        '127.0.0.1:3301',
        '127.0.0.1:3302',
    },
    replication_connect_timeout = 60,
}
box.schema.user.grant('guest', 'replication')

-- Step 4.
box.space._schema:insert{'asdf'}

-- Step 10.
-- Instance 2 is running after rebootstrap, but the applier on Instance 1
-- is still stopped with the original error.
box.info.replication[2].upstream
box.space._schema:get{'after_rebootstrap'}

-- Step 11.
-- After restarting replication, it works normally.
box.cfg{replication = {}}
box.cfg{replication = {
    '127.0.0.1:3301',
    '127.0.0.1:3302',
}}

box.space._schema:get{'after_rebootstrap'}


-- Instance 2.
-- Step 2.
box.cfg{
    listen = '127.0.0.1:3302',
    replication = {
        '127.0.0.1:3301',
        '127.0.0.1:3302',
    },
    instance_uuid = 'aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa',
    read_only = true,
    replication_connect_timeout = 60,
}

-- Step 3.
box.cfg{replication = ''}

-- Step 5.
box.cfg{read_only = false}
box.space._schema:insert{'asdf'}

-- Step 6.
os.exit()

-- Step 7. Wipe *.snap / *.xlog files of the replica, then restart it.

-- Step 8.
box.cfg{
    listen = '127.0.0.1:3302',
    replication = {
        '127.0.0.1:3301',
        '127.0.0.1:3302',
    },
    instance_uuid = 'aaaaaaaa-aaaa-aaaa-aaaa-aaaaaaaaaaaa',
    replication_connect_timeout = 60,
}

-- Step 9.
box.space._schema:insert{'after_rebootstrap'}

Start with the box.cfg replication setup and the applier state exposed by box.info.replication[2].upstream, then run the failed-replica wipe and restart sequence from the reproducer. Done means the applier on the other instance restarts automatically after rebootstrap and replication works in both directions without manually resetting box.cfg replication.