OF-3181: Allow static configuration of cluster node ID
Previously, an Openfire server generated a new cluster node ID on every startup, causing the identifier to change after each restart. This commit ensures that a server reuses the same node ID over time by persisting the initially generated value.
The node ID is now stored in `openfire.xml` under a new property: `clustering.nodeid`
- If the property does not exist at startup, a random node ID is generated and persisted.
- If the property exists, its value is reused.
This allows administrators to optionally define stable, human-recognizable node IDs. When doing so, administrators are responsible for ensuring uniqueness across all cluster members.
This change also removes the concept of a "default" node ID. Since Openfire 4.4.0 (OF-1705), node IDs no longer change when joining a cluster, making the distinction between default and runtime IDs obsolete. As a result, there is now a single node ID per server.
As a functional consequence, the node ID is no longer available until after server initialization (it is no longer a static field). This is expected to have little to no practical impact, as the previous static value was not publicly accessible.
OF-3164: Avoid user lookup for JIDs without node to reduce exception overhead
Prevent execution of user lookup logic when the provided JID has no node. JIDs without a node part cannot represent a local user, yet currently flow through UserManager.isRegisteredUser(), which relies on exception-based control flow when the user does not exist.
Thread dumps show significant CPU time spent constructing UserNotFoundException instances (Throwable.fillInStackTrace) on this hot path.
This change short-circuits the lookup for node-less JIDs, avoiding the exception-heavy path and reducing unnecessary CPU usage under load.
Note: This is a partial workaround. A full fix should refactor user existence checks to avoid exceptions for normal control flow.
