Why did the do this before?
In the pre-shader days, there were significant differences between OpenGL and D3D. There were also other APIs, like GLIDE. These have since fallen by the wayside.
If the drivers for some particular hardware didn't support OpenGL very well, it might support D3D, so you needed a D3D path. But you could do more stuff with OpenGL via extensions. So having a GL rendering pipeline was important for features.
Why aren't they still doing it?
Because there's no point anymore.
OpenGL and D3D are cosmetically different, but for the most part functionally equivalent. Outside of certain proprietary extensions, you can't really do more with either API.
And the sad fact is that D3D drivers are more likely to be stable on Windows than OpenGL drivers are. Intel's OpenGL drivers are somewhere between terrible and Godawful, while their D3D drivers are somewhere between crap and not entirely incompetent.
You are either writing a Windows-only application, or you aren't. If you are Windows-only, then the reasons for writing a GL rendering pipeline are... rather dubious. You will gain very little by it without using proprietary extensions, and even that won't buy you a lot these days.
And if you're writing a cross-platform or just non-Windows app, then OpenGL is the way to go. Of course, if you want to be cross-platform on Windows, then it wouldn't be a terrible idea to have a D3D rendering path. But at the same time, it isn't that bad to just rely on OpenGL, depending on how wide a net you want to cast for your application.