JCR-2288: [Cluster] Adding nodes to the same parent node causes TimeoutException Problem analysis - The goal of this task is to get rid or limit the risk of getting TimeoutException when we concurrently add nodes to the same parent node in a cluster environment.
Fix description - When we add a new child node, we have to invalidate the list of child nodes of the parent node in the cache which was done by a simple remove of the related cache entry. The problem with this approach is the fact that this remove was done in a global tx so when it is done concurrently we can face deadlock like in this particular use case. The fix simply isolates the invalidation by doing it outside the global tx in order to highly reduce the risk of such deadlocks. The second part of the fix ensures that if we do a local change, the related commands commit/rollback remain local because before the fix they were replicated even for local changes.