Add OpenMP barrier before releasing threadinfo & global communicator to avoid race (#225)

- Added `#pragma omp barrier` just before threads start releasing their threadinfo / global
  communicator.
- This ensures all threads reach this sync point, preventing interleaved cleanup.

Co-authored-by: harsdave <harsdave@amd.com>

frame/base/bli_sba.c

@@ -101,7 +101,7 @@ void* bli_sba_acquire
 	}
 #else
 
-	block = bli_malloc_user( req_size, &r_val );
+	block = bli_malloc_intl( req_size, &r_val );
 
 #endif
 
@@ -145,7 +145,7 @@ void bli_sba_release
 	}
 #else
 
-	bli_free_user( block );
+	bli_free_intl( block );
 
 #endif
 }

frame/thread/bli_l3_sup_decor_openmp.c

@@ -121,6 +121,8 @@ err_t bli_l3_sup_thread_decorator
 		  thread
 		);
 
+		/* Synchronize all threads */
+        #pragma omp barrier
 		// Free the current thread's thrinfo_t structure.
 		bli_l3_sup_thrinfo_free( rntm_p, thread );
 	}