holepunch: add multiaddress filter

[dennis-tra]

This PR adds a new option to the holepunch service that allows filtering of multi addresses during the DCUtR protocol.

[dennis-tra]

I'm inconsistently referring to the multiaddress filter as AddrFilter or just filter (see L16) 🤷‍♂️

[marten-seemann]

Code LGTM, but CI still seems unhappy.

[marten-seemann]

@dennis-tra Can you take a look at this?

[dennis-tra]

@marten-seemann I don't get why CI is so angry at me #worksonmymachine

There was indeed a failing test that I have fixed. Now, when I run the holepunch package tests in isolation all pass. After I did that I ran the whole test suite locally and I got the below error (same as CI). Then I ran the suite again and now it passes (as the results are cached I believe).

The error is:

      holepunch_test.go:236: 
          	Error Trace:	/home/runner/work/go-libp2p/go-libp2p/p2p/protocol/holepunch/holepunch_test.go:236
          	Error:      	"failed to open hole-punching stream: protocol not supported" does not contain "aborting hole punch initiation as we have no public address"
          	Test:       	TestFailuresOnInitiator/no_addrs_after_filtering

Could it be that there's an other interfering test? Perhaps a mocked host not closed?

[marten-seemann]

Now, when I run the holepunch package tests in isolation all pass. After I did that I ran the whole test suite locally and I got the below error (same as CI). Then I ran the suite again and now it passes (as the results are cached I believe).

That sounds like a flaky test. Go runs the tests from different packages in separate binaries, so it shouldn't be possible that tests from another package influence your holepunching tests.

[marten-seemann]

@dennis-tra Looks like there was a race condition in the test. I added a fix in dennis-tra#1, but since this is in your fork, I can't apply it here. Can you please that PR?

[dennis-tra]

@marten-seemann CI seems to be happier - but not quite super happy. Some crypto test is still failing:

=== RUN   TestPreferRelayV2
2022/11/17 10:35:16 failed to sufficiently increase receive buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/lucas-clemente/quic-go/wiki/UDP-Receive-Buffer-Size for details.
    autorelay_test.go:192: 
        	Error Trace:	/home/runner/work/go-libp2p/go-libp2p/p2p/host/autorelay/autorelay_test.go:192
        	Error:      	Condition never satisfied
        	Test:       	TestPreferRelayV2
--- FAIL: TestPreferRelayV2 (3.03s)

Seems to be an unrelated test. Would it suffice to rerun the Go Cryto test and hope that it then works? 🤞