Simplifying handling of `node_name` vs. `hostname` vs. `opensearch_network_host`

following assumptions/thoughts:
- we either want to bind the rest api and opensearch to all available interfaces or to the same, single ethernet interface
- in this case, we can remove `opensearch_network_host` from the config. if `http_bind_address` contains sth. that is not the default, we use the host part of this also for the opensearch `network.host` option

- if `http_bind_address` is set to sth differing from the default, we can use the hostname part also for the hostname and for the nodename config parameters, because we assume that this is also the dns resolved name we want to use to connect machines

- if `http_bind_address` is set to the default, we use the the hostname from `Tools.getCanonicalHostname()` also as the nodename in the config unless nodename is set explicitly to sth else

- within docker, if nothing is set in the config file, using `Tools.getCanonicalHostname()` works to get a running system

- if all this holds true, we drop/not merge my PR (https://github.com/Graylog2/graylog2-server/pull/16572) to use the node-id short format as the nodename