Use lazy-regex crate instead of plain regex

[LiamPattinson]

Currently, the tree-sitter based rules start at the root node of a file and scan the whole tree to find any matches. They then sometimes use regex to inspect what they find further. Any regular expressions used are therefore compiled once per file.

Inverting the order of operations as suggested by #26 would mean having to compile a regular expression at each node, which would be terrible for performance.

This PR pre-emptively solves this problem by replacing all instances of the plain Regex crate with lazy-regex, which provides similar capabilities but compiles only once per run.