Fix EntryForRegistry code coverage randomization

[tianon]

I hate this, but golang/go#61900 might someday make it slightly cleaner.

See #32 (comment):

Because our iteration order is random, sometimes we get coverage on this first continue and sometimes we do not! (and the function is small enough that this actually affects the overall percentage by like 9.1%!)

[tianon]

Nice, and because "" sorts above everything else, we end up with the "good" case where the continue is included in our coverage. 👍

[tianon]

github.com/docker-library/meta-scripts/registry/client.go:135: EntryForRegistry 81.8% (main)

vs

github.com/docker-library/meta-scripts/registry/client.go:136: EntryForRegistry 86.7% (this PR)

[tianon]

I think it's only a ~5% increase (instead of ~9%) because of all the lines this adds lolsob