If you're having trouble because generating a 3 way string compare in Go isn't fast enough and delegating to bytes is hard to get correct, my recommendation would be to file an issue with Go. It's not difficult for them to make strings.Compare faster, and I imagine real world evidence that it'd be substantially helpful would be enough to motivate it.

Though, if 3 way compare is the bottleneck, it seems like a real possibility that there are algorithmic optimizations that may be more effective than a faster 3 way compare, like using native comparison operators, hashing, using byte slices, etc. Or maybe not; I don't think I've seen a case where it's a bottleneck so I might be inaccurately imagining the cases where it'd happen.