fix(kademlia): align DHT message and provider limits for interop

[richard-ramos]

Summary

Raises the Kademlia RPC frame limit from 4 KiB to 4 MiB so larger /ipfs/kad/1.0.0 responses, such as peer lists with multiple multiaddrs, are not rejected before decoding.

Also updates the default provider record expiration interval to 48 hours and keeps stored value records independently bounded by DefaultMaxValueSize.

Service discovery now uses its own ServiceDiscoveryMaxMsgSize constant, preserving the existing 4 KiB limit for those extension messages instead of inheriting the larger Kademlia RPC cap.

Affected Areas

• Peer Management / Discovery

  • Kademlia provider record expiration defaults
  • Kademlia RPC message size limit

• Protocol Logic

  • Service discovery frame-size handling remains explicitly capped at 4 KiB

Impact on Library Users

Kademlia nodes can accept larger DHT replies from go-libp2p DHT peers. Provider records are retained for 48 hours by default instead of 30 minutes.

No API migration is required for existing users.

Risk Assessment

The larger Kademlia read cap increases the maximum accepted RPC frame size. Stored value records remain separately limited by DefaultMaxValueSize, and service discovery keeps its previous 4 KiB cap.

References

• go-libp2p-kad-dht Amino defaults: https://github.com/libp2p/go-libp2p-kad-dht/blob/master/amino/defaults.go