fix(resolve): correctly handle .eth domains

[Stebalien]

This should have been handled down inside the DNSLink resolver. Otherwise, we'll
break any name happens to end in .eth.

also fixes #6699

[aschmahmann]

@Stebalien can you clarify if my understanding is correct.

The fix to #6699 is to use the updated go-is-domain library that accounts for ETH as a tld extension which deals with a bug at https://github.com/ipfs/go-ipfs/blob/d9f2bafa45693db2388544097089b3c7a5182110/core/corehttp/ipns_hostname.go#L27

Moving the .eth -> .eth.link code into the DNSLink resolver from the namesys resolver is for sanity/correctness, but is unrelated to #6699.

If the above is correct why are the tests being removed, the bug was in the HTTP API not in Namesys?

we'll break any name happens to end in .eth

Sure, although we don't currently have any resolver other than DNSLink that can end in .eth, right?

[Stebalien]

Moving the .eth -> .eth.link code into the DNSLink resolver from the namesys resolver is for sanity/correctness, but is unrelated to #6699.

Yes.

If the above is correct why are the tests being removed, the bug was in the HTTP API not in Namesys?

Those tests used a fake DNSLink resolver (bypassing the real dnslink resolver). However, thanks for calling me on that. I've added the tests to the real DNSLink resolver's test suite.

[Stebalien]

Sure, although we don't currently have any resolver other than DNSLink that can end in .eth, right?

/ipfs/QmFoo/foo.eth? I think?

[Stebalien]

Ok, the API tests are stalling on master: #6703