New tokenize-complete hook

[RunDevelopment]

Motivation
Some languages have custom tokenizers implemented as hooks (e.g. JSX). Right now, they all use the after-tokenize hook. This is a problem because components like Markup Templating don't run this hook (because they can't, see the implementation).

Therefore, we need a new hook that gets run before after-tokenize and inside the Prism.tokenize function itself.

Description
Since the hook is supposed to be called inside Prism.tokenize, we need to change the function a little. It needs to take a new optional argument: the id of the grammar (the language string).

Only if this argument is present, will the hook be run. This ensures that the hook is only run for top-level grammars and not inside grammars.

Kinda like this:

function tokenize(text: string, grammar: Grammar, language?: string): TokenStream {
	// ...
	matchGrammar(text, tokenList, grammar, tokenList.head, 0);

	var tokens = toArray(tokenList);

	if (language) {
		var env = { tokens, language };
		Prism.hooks.run('tokenize-complete', env);
		return env.tokens;
	} else {
		return tokens;
	}
}

This approach will also make it easy to implement language-specific parsers.

---

Together with #2948, these languages specific parsers will also work as embedded languages.

[RunDevelopment]

Closed in favor of #3539.