refactor(vary): use word character inside regex character class

[Fdawgs]

• Replaced a-zA-Z0-9_ inside character class with \w word character, which is equivilent
• Added the u unicode flag to every regex pattern, which acts as a strict mode for regex. See perf(index): convert unused capture group to non-capture group fastify-autoload#330

Checklist

• run npm run test and npm run benchmark
• commit message and code follows the Developer's Certification of Origin
  and the Code of conduct

[Fdawgs]

Throwing this into a draft for now. @Uzlopak I noticed you've been making regex improvements in Node itself (nodejs/node#49643), do you know if using a \w character over [a-zA-Z0-9_] impacts performance at all?

[Uzlopak]

Tbh i never benched it. And i have no clue if it could have an impact. I mean theoretically \w gets just translated to the specific chars internally. But maybe the regex engine optimizes it somehow.

But on the other hand i never used \w because like 10 years ato I always needed to lookup if i am not mistaking it with \W :))).

[Fdawgs]

Just used the benchmark in benchmark/vary.js

master:

PR:

[Uzlopak]

LGTM

[mcollina]

lgtm