It looks safer to return C_OK from freeMemoryIfNeeded() when clients are
paused because returning C_ERR may prevent success of writes. It is
possible that there is no difference in practice since clients cannot
execute writes while clients are paused, but it looks more correct this
way, at least conceptually.

Related to PR #4028.

1. brpop last key index, thus checking all keys for slots.
2. Memory leak in clusterRedirectBlockedClientIfNeeded.
3. Remove while loop in clusterRedirectBlockedClientIfNeeded.

Proposed fix to https://github.com/antirez/redis/issues/4027

Fix a few typos/adjust wording in `create-cluster` README

commands.

Close #3993.

This avoids Helgrind complaining, but we are actually not using
atomicGet() to get the unixtime value for now: too many places where it
is used and given tha time_t is word-sized it should be safe in all the
archs we support as it is.

On the other hand, Helgrind, when Redis is compiled with "make helgrind"
in order to force the __sync macros, will detect the write in
updateCachedTime() as a read (because atomic functions are used) and
will not complain about races.

This commit also includes minor refactoring of mutex initializations and
a "helgrind" target in the Makefile.

The __sync builtin can be correctly detected by Helgrind so to force it
is useful for testing. The API in the INFO output can be useful for
debugging after problems are reported.

Only useful for when no atomic builtins are available.

More work to do with server.unixtime and similar. Need to write Helgrind
suppression file in order to suppress the valse positives.

Instead of giving the module background operations just a small time to
run in the beforeSleep() function, we can have the lock released for all
the time we are blocked in the multiplexing syscall.