Pure-Rust core for words2num2 — the inverse of num2words2.

Why this is small

119 of words2num2's 120 locales never had a hand-written parser. They use Words2Num_Base, which materialises a reverse lookup table by calling num2words across LOOKUP_RANGE (range(-1, 10001)) — 10,002 renders — and then does a dict hit. Only en is hand-written.

So the port is: the generic table backend + _normalize + the en grammar parser. The table is built by calling the Rust num2words core (num2words2-core) directly, which is where the speedup comes from.

This crate has no PyO3 dependency: it is the pure-Rust engine. The words2num2-py crate is a thin PyO3 binder over the public API here.