WSREP_ON is unnecessarily expensive to evaluate

The macro WSREP_ON used to be defined simply as an alias the global Boolean variable global_system_variables.wsrep_on. A more complex variant that invokes strcmp was introduced, in MDEV-16799 renamed to WSREP_ON_NEW, and related to MDEV-16405 (Galera 4) in 10.4.2 the more efficient macro WSREP_ON was replaced altogether.

The following looks grossly inefficient:

#define WSREP_ON                         \

  ((global_system_variables.wsrep_on) && \

   wsrep_provider                     && \

   strcmp(wsrep_provider, WSREP_NONE))

/* use xxxxxx_NNULL macros when thd pointer is guaranteed to be non-null to

 * avoid compiler warnings (GCC 6 and later) */

#define WSREP_NNULL(thd) \

  (thd->variables.wsrep_on && WSREP_ON)

#define WSREP(thd) \

  (thd && WSREP_NNULL(thd))

Why does the Boolean predicate WSREP_ON have to be so complex? We should be able to compute one global Boolean variable, something like this:

extern my_bool WSREP_ON;

#define WSREP_NNULL(thd) (WSREP_ON && thd->variables.wsrep_on)

#define WSREP(thd) (thd && WSREP_NNULL(thd))

The Boolean variable WSREP_ON would be updated by the update callback functions that would be attached to global_system_variables.wsrep_on and wsrep_provider.

There are rather many checks for WSREP_ON, so this could affect performance, especially starting with MariaDB Server 10.4. I feel that this is important, because the code is being invoked even when Galera is not being used.

MDEV-7962 was filed for performance regressions due to a similar check that is implemented in a non-inlined function wsrep_on(). Could we replace that one with an efficient implementation of the macro WSREP_ON?