feat: AutoTLS with AutoNATv2 client

[sukunrt]

This updates libp2p and autotls to test changes autonatv2 changes. Post this we'd not need to have a large registration delay.
The corresponding autotls change is here:

• fix: use autonatv2 event to begin cert management ipshipyard/p2p-forge#70
• Deployment plan for AutoNAT v2 #10091

[lidel]

Thank you @sukunrt, fixed CI and took this for a spin and it seems to work as expected.

Two asks:

• One CI test failure remains: TestDHTConnectivityFast from ./test/integration/wan_lan_dht_test.go started failing
  • I've re-run it and it fails every time. I'm not familiar with it, it was introduced in 2020 before my time in go-ipfs (#7151). Mind taking a look?
• Is this PR effectively disabling AutoNAT v1 client or is go-libp2p node running both? p2p-forge/client seems to support both, which is good for transition period, but unclear if both events are emitted by go-libp2p v0.42.0-alpha
  • Should we update docs at https://github.com/ipfs/kubo/blob/master/docs/config.md#autonatservicemode and code in ./core/node/libp2p/nat.go to remove legacy-v1 and AutoNATServiceEnabledV1Only, or ok to keep for one more release?
    • tldr: we have both for now, so keeping as-is

[sukunrt]

Is this PR effectively disabling AutoNAT v1 client or is go-libp2p node running both? p2p-forge/client seems to support both, which is good for transition period, but unclear if both events are emitted by go-libp2p v0.42.0-alpha

go-libp2p v0.42 supports both and depends on both. Autorelay still uses autonatv1.
go-libp2p v0.43 will stop using autonatv1 internally. I don't mind supporting autonatv1 for a couple more releases, but users can safely stop using it.

So kubo can remove autonatv1 from go-libp2p v0.43 onwards.

[sukunrt]

One CI test failure remains: TestDHTConnectivityFast from ./test/integration/wan_lan_dht_test.go started failing
I've re-run it and it fails every time. I'm not familiar with it, it was introduced in 2020 before my time in go-ipfs (#7151). Mind taking a look?

This is a mocknet bug. The new addrs manager depends on the new listen addrs event from the swarm. Fixed mocknet to send this event and to use mocknet.Listen instead of peerstore.AddAddr to add a new address to the host.

[lidel]

Thank you @sukunrt!

I've switched this PR to commit from main branch at https://github.com/ipshipyard/p2p-forge with your fix.

On top of green CI, I did basic manual regression/smoke-tests and AutoTLS works as expected for IPv4, I'm going to merge this so we can include it in RC1.

Once we have final go-libp2p 0.42, we can bump go.mod before final Kubo 0.36.