context size vs speed in GGUF with vLLM

[cudacrate]

I'd start with the simpler stack.

Did a small comparison this afternoon. I couldn't reproduce the slowdown until I pushed context much higher. I expected the model choice to matter more than it did.