Update publicsuffix-go to 6f3c5059

[weppos]

Sorry for the extra noise, but thanks to external contributions the lookup has been drastically improved and I've been working with some consumers to double check everything is working as expected.

Despite I carefully inspected the changes, I want to make sure no subtle use cases are broken.

If you are interested in, here's the benchmarks:
weppos/publicsuffix-go#133 (comment)

Before:

➜  publicsuffix-go git:(master) ✗ GOCACHE=off  go test ./... -bench=.
?   	github.com/weppos/publicsuffix-go/cmd/load	[no test files]
PASS
ok  	github.com/weppos/publicsuffix-go/net/publicsuffix	0.019s
goos: darwin
goarch: amd64
pkg: github.com/weppos/publicsuffix-go/publicsuffix
BenchmarkDomain-4   	    1000	   1522570 ns/op
BenchmarkXNet-4     	  500000	      3267 ns/op
PASS
ok  	github.com/weppos/publicsuffix-go/publicsuffix	3.383s

After

➜  publicsuffix-go git:(master) GOCACHE=off  go test ./... -bench=.
?   	github.com/weppos/publicsuffix-go/cmd/load	[no test files]
PASS
ok  	github.com/weppos/publicsuffix-go/net/publicsuffix	0.017s
goos: darwin
goarch: amd64
pkg: github.com/weppos/publicsuffix-go/publicsuffix
BenchmarkDomain-4   	  500000	      3140 ns/op
BenchmarkXNet-4     	  500000	      3313 ns/op
PASS
ok  	github.com/weppos/publicsuffix-go/publicsuffix	3.317s

Performance are now equivalent to the golang/x package, still providing the flexibility that was the core of the design of my alternate version.

If you want to inspect the changes, the PR is weppos/publicsuffix-go#143

---

As per contribution rule, tests are green:

➜  publicsuffix-go git:(master) GOCACHE=off  go test ./...
?   	github.com/weppos/publicsuffix-go/cmd/load	[no test files]
ok  	github.com/weppos/publicsuffix-go/net/publicsuffix	0.021s
ok  	github.com/weppos/publicsuffix-go/publicsuffix	0.021s

[weppos]

I can see the specs failed.

--- FAIL: TestExtractDomainIANASuffix_Valid (0.00s)
	pa_test.go:446: "kobe.jp": got "kobe.jp", want "jp"
FAIL
FAIL	github.com/letsencrypt/boulder/policy	0.038s

I'm looking into the case.

[weppos]

There was an edge case that I was able to discover thanks to your test suite. I fixed the issue in weppos/publicsuffix-go@f8afde6, added regression tests to the lib test suite, and now also LE tests are green.

[cpu]

Thanks @weppos, I really appreciate that you found the time to backport this update to Boulder. Nice to hear our unit tests were useful as well!!

I scanned the diff on the upstream library and it looks good to me.