[v2] Match tokens via named capturing groups

Prism's token syntax is helpful for certain things, but for other use cases a single regex with named capturing groups can be easier.
I propose we just use all named capturing groups as token names (automatically converted to kebab-case)
If that becomes a problem, we can introduce a setting to ignore some of them.

This can be combined with regular tokens, they'd just operate on the unmatched text. The tokens matched in the regex can also be specified again in the literal if we want to specify inside or $language. E.g. imagine something like this for the markdown code blocks:

'code-block': {
	pattern: /^```(?<codeLanguage>[a-z-]+).+(?:\n|\r\n?)(?<code>[\s\S]*)?(?:\n|\r\n?)```$/i,
	'punctuation': /```/,
	'code': {
		$language: groups => groups.codeLanguage,
	}
}

Note that this means inside can also be a regex now.

Here's a more substantial example: https://github.com/PrismJS/prism/commit/586bd55e5661a12f9876cb972761f317a5314305

Start by reviewing the issue's Markdown code-block example and commit 586bd55e5661a12f9876cb972761f317a5314305. Define the behavior for converting all named capture groups to kebab-case token names, combining them with regular tokens, and allowing regex-based inside rules and group references; completion requires these cases to work together.