test: wait until cluster leaders are rw

Tarantool 3.0 instances start in ro mode. It is possible that masters
are not available for write requests right after test case has been
started.

Cartridge servers run only on Tarantool 1.x and 2.x, so they are
expected to work fine without additional changes: Cartridge helpers
checks are sophisticated enough.

Part of #412
Part of #415

test/integration/select_readview_test.lua

@@ -2307,6 +2307,8 @@ pgroup.test_stop_select = function(g)
     g.cluster:server('s2-master'):start()
 
     if g.params.backend == helpers.backend.VSHARD then
+        g.cluster:server('s2-master'):wait_for_rw()
+
         local bootstrap_key
         if type(g.params.backend_cfg) == 'table'
         and g.params.backend_cfg.identification_mode == 'name_as_key' then

test/vshard_helpers/server.lua

@@ -114,6 +114,16 @@ function Server:wait_for_readiness()
     end)
 end
 
+function Server:wait_for_rw()
+    return wait_cond('rw', self, function()
+        local ok, is_ready = pcall(function()
+            self:connect_net_box()
+            return self.net_box:eval('return box.info.ro') == false
+        end)
+        return ok and is_ready
+    end)
+end
+
 function Server:wait_election_leader()
     -- Include read-only property too because if an instance is a leader, it
     -- does not mean it finished the synchro queue ownership transition. It is

test/vshard_helpers/vtest.lua

@@ -497,6 +497,7 @@ local function cluster_new(g, cfg)
 
     for _, master in pairs(masters) do
         master:wait_for_readiness()
+        master:wait_for_rw()
         master:exec(function(cfg)
             -- Logged in as guest with 'super' access rights. Yet 'super' is not
             -- enough to grant 'replication' privilege. The simplest way - login