litep2p/discovery: Fix memory leak in litep2p.public_addresses()

[lexnv]

This PR ensures that the litep2p.public_addresses() never grows indefinitely.

• effectively fixes subtle memory leaks
• fixes authority DHT records being dropped due to size limits being exceeded
• provides a healthier subset of public addresses to the /identify protocol

This PR adds a new ExternalAddressExpired event to the litep2p/discovery process.

Substrate uses an LRU address_confirmations bounded to 32 address entries.
The oldest entry is propagated via the ExternalAddressExpired event when a new address is added to the list (if capacity is exceeded).

The expired address is then removed from the litep2p.public_addresses(), effectively limiting its size to 32 entries (the size of address_confirmations LRU).

cc @paritytech/networking @alexggh

[dmitry-markin]

Why do we have so many (>32) external addresses confirmed? Is there a linked issue in polkadot-sdk?

[lexnv]

Why do we have so many (>32) external addresses confirmed? Is there a linked issue in polkadot-sdk?

This issue was reproduced in Rococo on litep2p validators, don't believe we have an gitissue for it. It looks like over time peers report via identify protocol connections on different ports.
Libp2p has a few events on which we clean up the addresses (AddressExpired / ExpiredListenAddr / ListenerClosed / port-reuse logic).

[alexggh]

Why do we have so many (>32) external addresses confirmed? Is there a linked issue in polkadot-sdk?

This is the long standing issue for this happening on rococo: #3519 (comment), we never uncovered why so many external addresses accumulate on rococo.

[paritytech-cmd-bot-polkadot-sdk]

Created backport PR for stable2407:

• [stable2407] Backport #5998 #6083 with remaining conflicts!

Please cherry-pick the changes locally and resolve any conflicts.

git fetch origin backport-5998-to-stable2407
git worktree add --checkout .worktree/backport-5998-to-stable2407 backport-5998-to-stable2407
cd .worktree/backport-5998-to-stable2407
git reset --hard HEAD^
git cherry-pick -x 38aa4b755c5bb332436950363a9826a91de4b2ef
git push --force-with-lease

[paritytech-cmd-bot-polkadot-sdk]

Successfully created backport PR for stable2409:

• [stable2409] Backport #5998 #6084

[dmitry-markin]

Why do we have so many (>32) external addresses confirmed? Is there a linked issue in polkadot-sdk?

This is the long standing issue for this happening on rococo: #3519 (comment), we never uncovered why so many external addresses accumulate on rococo.

In litep2p backend we have the requirement that at least 2 different peers should report the same external address for it to become confirmed. So, in theory, it should be better with litep2p backend.