feat: dead peer detection

[maqi]

This PR contains the work of detection dead peers and remove it from the kad.

Note: this relies on flying requests among nodes. i.e. the detection only works when there is data/requests being exchanged among the network.
A quiet network, say startup a network of 25 nodes locally, then kill one of it, won't trigger the detection.

[reviewpad]

AI-Generated Summary: This pull request introduces dead peer detection for the network module. The changes include adding a dependency on lru_time_cache and updating Cargo.lock and safenode/Cargo.toml accordingly. In addition, new code has been added to safenode/src/network/event.rs and safenode/src/network/mod.rs to detect dead peers, remove them from the list of known peers and handle potential outgoing connection errors. The SwarmDriver struct now also includes an LruCache for keeping track of potential dead peers.

[reviewpad]

Reviewpad Report

⚠️ Warnings

• Please link an issue to the pull request

[reviewpad]

AI-Generated Summary: This pull request includes a new feature for dead peer detection in the safenode's networking module. The changes include 42 new lines of code, which modifies 4 files: Cargo.lock, safenode/Cargo.toml, safenode/src/network/event.rs, and safenode/src/network/mod.rs. Among these changes, the LruCache dependency has been added, which is used to maintain a cache for potential dead peers. This new dead peer detection leads to faster get_closest_peers results and ensures a corrected partially targeted replication.