MXS-2342 Run MariaDBMonitor diagnostics concurrent with the monitor loop

This fixes some situations where MaxAdmin/MaxCtrl would block and wait
until a monitor operation or tick is complete. This also fixes a deadlock
caused by calling monitor diagnostics inside a monitor script.

Concurrency is enabled by adding one mutex per server object to protect
array-like fields from concurrent reading/writing.

server/modules/monitor/mariadbmon/mariadbserver.hh

@@ -15,6 +15,7 @@
 #include <functional>
 #include <string>
 #include <memory>
+#include <mutex>
 #include <maxscale/monitor.h>
 #include <maxbase/stopwatch.hh>
 #include "server_utils.hh"
@@ -114,6 +115,7 @@ public:
         bool log_slave_updates = false; /* Does the slave write replicated events to binlog? */
     };
 
+
     /* Monitored server base class/struct. MariaDBServer does not own the struct, it is not freed
      * (or connection closed) when a MariaDBServer is destroyed. */
     MXS_MONITORED_SERVER* m_server_base = NULL;
@@ -548,6 +550,12 @@ private:
         DISABLE
     };
 
+    /* Protects array-like fields from concurrent access. This is only required for fields which can be
+     * read from another thread while the monitor is running. In practice, these are fields read during
+     * diagnostics-methods. Reading inside monitor thread does not need to be mutexed, as outside threads
+     * only read the values. */
+    mutable std::mutex m_arraylock;
+
     bool               update_slave_status(std::string* errmsg_out = NULL);
     bool               sstatus_array_topology_equal(const SlaveStatusArray& new_slave_status);
     const SlaveStatus* sstatus_find_previous_row(const SlaveStatus& new_row, size_t guess);