#3174 Move initialization to (a pointer to) EMSG at the Cglobals_get() level (when Cglobals_get returns > 0)

[bugs #2931] side effect of longname version on repack : Use stage_util_max_stcp_per_request() instead of hardcoded 1000 value
Added a missing print on stderr in case of error
Changed a print from stdout to stderr

[bugs #2931] side effect of longname version on repack : Added stage_util_max_stcp_per_request() and stage_util_max_stpp_per_request() functions

mount read-only flag.

Remove explicit prototype of arguments for functions _Cuuid_write_state() and _Cuuid_format_uuid_v1(). Compiler complains$
#if (defined(hpux) || defined(sun))
/* No prototype */
#else
/* Full prototype */
#endif

usleep() between 0 and 1 second

In Cpool_assign(), condider that it is a fatal error if it cannot assign a thread that was pre-determined by Cpool_next_index()
In Cpool_assign(), always cleanup previous call to Cpool_next_index()

Fix failover code inside Cpool_next_index_timeout() that was setting current->flag to 0 instead of -2 (parent not waiting)

Removed a branch in Cpool_next_index that checked if we were returning the same thread as a previous call : it is legal to not get twice the same index when calling twice Cpool_next_index

Makes sure that consecutive calls to Cpool_assign() without Cpool_next_index() in between are okay : predicate current->flag now have the folwlowing meaning:
-2 : parent not waiting
-1 : parent waiting, then the child that answered puts its index (=>0) into

Added timeout in the list of parameters in a log inside _Cpool_nextindex_timeout_ext()
Suppressed an #ifdef/#endif CPOOL_DEBUG pait that was causing Cthread_mutex_unlock_ext(current->state_cthread_structure[i]) to NOT be compiled with Cpool.c was compiled without CPOOL_DEBUG -> very quick deadlock

Respect timeout when calling for a lock on current->state_cthread_structure[child_number] when a previous call to Cpool_next_index() was successful and returned _that_ child_number. If not timeout given to Cpool_assign(), then there will be no timeout to the lock

Always maintain a lock on current->state_cthread_structure[index] in the child, releasing atomically only when doing condition wait on it
Always use trylock() on it in Cpool_next_index()
Always use lock() on it if Cpool_next_index() already returned that index trylock() otherwise
In routines Cpool_next_index() and Cpool_assign() always own a lock on lock_parent_cthread_structure, releasing atomically only when doing condition wait on it, then a child will be able to take over and send a signal

Typo nb_thread -> nb_pool

Removed global lock lock_child_cthread_structure in favour of per-thread lock (on every thread's state)
Optimized Cpool_assign() when it is preceeded by a successful call to Cpool_next_index()