Enable eyecite with vectorscan in prod (instead of using Python's re module)

[mlissner]

Amazingly, in prod we're not using the CPU-optimized version of eyecite, which uses vectorscan:

https://github.com/VectorCamp/vectorscan

We should do that. It's faster! It saves energy! Let's do that. I think we just need to enable/install a dependency, but the eyecite docs should have the details.

[mlissner]

Over in freelawproject/eyecite#177, we bumped to hyperscan 0.7.7, which switches to vectorscan and thereby enables support for ARM processors. That might be enough to make vectorscan work, but I'd like to be sure we're using it.

[mlissner]

Yeah, looks like we need to enable the correct tokenizer: https://github.com/freelawproject/eyecite/?tab=readme-ov-file#tokenizers