Assign opensearch node attribute to tie datanode instance with opensearch instance

[todvora]

Bind datanodes opensearch health check to specific opensearch process instance. It may happen that the management of the process fails but the underlying opensearch still runs fine. Then a watchdog restarts the process and meanwhile the healtcheck sees a healthy opensearch. This opensearch is not the currently managed one. It then confuses the watchdog, restarting its error counters, causing infinite restart loop.

[todvora]

Maybe we don't need any new APIs or attributes and we can simply compare the pid opensearch gives us with the pid we hold?

[moesterheld]

@todvora sounds reasonable. that should work