Ability to group multiple features for the powerset

[jhpratt]

With an increasing number of features, the powerset inherently grows exponentially. As such, even the ability to group two features (dependencies in my case) together would provide a significant improvement in CI time. The use-case behind this is for features that are wholly unrelated and do not interact in any way. In my case, I have serde and rand implementations, but enabling one would never cause the other to fail. Because of this, having an option like --group-features serde,rand would be great. The flag could be passed multiple times to create multiple groups.