Add tests to ensure semver leading 'v' doesn't mess us up

[sdboyer]

When a constraint is specified with a leading v, but the underlying version lacks it, or vice versa, it's correct to elide that difference and have both cases match.

I don't think we do this now, but I'm honesty not sure. Need some tests to check.

[sdboyer]

At least half of the importance here is just clearly identifying where in the solving algorithm the decision gets made about what form we actually use. I'm pretty sure it takes from the vcs when populating the solution, so we should always be good as long as the constraint Match()es the version, but being really clear about this would be a plus.

[fabulous-gopher]

This issue was moved to golang/dep#430