Duplicate releases of "same" version affects metrics

[foolip]

https://github.com/mdittmer/web-apis/pull/83 added data for a version of Safari 11.0, and it turns out that affects some of the graphs. It's the most clear in the API count graph.

Before:

After:

Following the links for the point of difference shows that before, "Safari 10.1.2 OSX 10.12.6" was being compared to "Safari 11.0.3 OSX 10.13.3". After adding the new data, "Safari 11.0 OSX 10.12.6" is compared to "Safari 11.0.3 OSX 10.13.3" instead.

In order to fix this, either duplicate releases should be deleted, or one would have to filter releases before producing metrics.