extract regex

[gaiarajan]

Overview

Extracts all regex so we only have to initialize it once.

Closes #410

Benchmark, 20,000 calls on a sample address:

tokenize():
Before: 7.16 µs/call
After: 5.00 µs/call (30% faster)

tag():
Before: 247.4 µs/call
After: 228.9 µs/call (7.5% faster)

I think this is a large enough performance improvement to be worth it.

pytest is still green (4972 / 4972 passing).