use fancy-regex instead of onig as tokenizers regex library (#172)

The version of Oniguruma used in `onig_sys` doesn't build on GCC 15 and
the oniguruma project itself got archived last week, so this PR switches
tokenizers to the fancy-regex backend.

`fancy-regex` also requires flipping on the `unstable_wasm` feature
until huggingface/tokenizers#1772 lands, that flag doesn't have any ill
effects though since everything WASM related downstream is behind
`target_arch` checks.

**tl;dr**: This fixes builds on Linux distros with newer GCC versions
like Arch Linux and Fedora.

Author: Slowki

Cargo.lock

@@ -11,5 +11,8 @@ toktrie = { workspace = true }
 serde = { version = "1.0.217", features = ["derive"] }
 serde_json = "1.0.138"
 anyhow = "1.0.95"
-tokenizers = { version = ">=0.20.0, <1.0.0", default-features = false, features = ["onig"] }
+tokenizers = { version = ">=0.20.0, <1.0.0", default-features = false, features = [
+    "unstable_wasm",
+    "fancy-regex",
+] }
 log = "0.4.25"