Create Rust-only index construction function

[brandonwillard]

We need a pure Rust function—call it build_regex_index—that only takes a regex string and a vocabulary or tokenizer object and (possibly) returns a HashMap version of the current Python index.

This is a step towards #1.

There are few things that need to be converted in order to do this:

• Determine the exact form of the return value.
  What works best for returning a Python object and future Rust-only use (i.e. Rust at run-time)?
• Determine if a vocabulary or tokenizer is better as input.
  Do we need to allow both?
  We eventually want tighter integration with tokenizers; should we start there? Also, could we get a Rust-based tokenizer from a Python-created transformers tokenizer object (that uses tokenizers under the hood, of course)?
• Handle regex-to-FSM conversion in Rust #10
  Start by finding a suitable Rust crate for this functionality (e.g. regex-automata).
• Convert reduced_vocabulary.