If you're open to writing your own, I found https://fgiesen.wordpress.com/2012/02/21/half-edge-based-mes... and https://fgiesen.wordpress.com/2012/03/24/half-edge-based-mes... to be good resources.
But it's performance I'm after. I have three options: 1. Burst compilation 2. Offload to the GPU (not sure if this is even a good use case for GPU compute) 3. Find a native library.
(3) is potentially a quick fix but might limit my ability to add functionality. But it also offers the tempting possibility of getting a ton of functionality for "free".