fix(kad): improve memory allocation when iterating over kbuckets

[stormshield-frb]

Description

Proposal to fix #5712.

I have changed to ClosestIter structure to only allocate when kbucket_size is higher than K_VALUE and only once along the life of ClosestIter. I think I did not break anything but I would really like some experienced people with Kademlia to take a look (@guillaumemichel 😉).

Notes & open questions

I have kept the sort_by in the Iterator implementation of ClosestIter but I think we could remove every allocation whatsoever if we used min_by but I'm not sure that it would not change the research behaviour that's why I did not changed it. If experienced developers think that it would be equivalent I will do it.
It was a dumb question. I did not fully understood how the ClosestIter was working.

Change checklist

• I have performed a self-review of my own code
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• A changelog entry has been made in the appropriate crates

[elenaf9]

@guillaumemichel do you have some time to review this? I am unfortunately not so familiar with kademlia.

[guillaumemichel]

Yes, I will try to review before the end of week.

[vbhattaccmu]

Hi @guillaumemichel do we have any estimate when this branch will get merged? Currently we are running into memory growth on our p2p light clients as well and are using a version of libp2p after this change was introduced. Having this change on latest commit will help us a lot. Thanks!