speed up read DiscoveryNode from network by skip

[fudongyingluck]

Description

In my cpu profiler of 7.10.2 version, the DiscoveryNode constructor occupied some percentage, as follow:

Although some optimizations have been done after this, some parts are still there.
But after PR #77991, most of time we actually do not use any information in the DiscoveryNode.
related to #94351

[DaveCTurner]

There is certainly room for improvement in the allocation traffic that happens when bringing a large cluster up from cold (see e.g. #81081) but I think it would be better to work in that direction rather than seeking more small tactical fixes like the one suggested here. Unfortunately this not a high-priority area of work for us at the moment: fully-restarting large clusters is just not something that happens often enough to warrant the effort needed.

[elasticsearchmachine]

Pinging @elastic/es-distributed (Team:Distributed)