Allow parsing mixed HTML and PHP

[sirbrillig]

When mixing HTML and PHP, we must ignore the HTML tokens and the closing PHP tokens. This PR alters the parsing to consider them "empty" for the sake of all the functions that skip over empty space. Notably, this allows considering a closing HTML tag as the end of the global scope.

Fixes #232

[jrfnl]

Suggested change

	return array_merge(
	array_values(Tokens::$emptyTokens),
	[
	T_INLINE_HTML,
	T_CLOSE_TAG,
	]
	);
	$empty = Tokens::$emptyTokens;
	$empty[T_INLINE_HTML] = T_INLINE_HTML;
	$empty[T_CLOSE_TAG] = T_CLOSE_TAG;
	return $empty;

Array functions like array_merge() and array_values() are "expensive".

Changing it to the above will be more efficient performance + memory-use-wise.

You could even choose to optimize it further like this:

Suggested change

	return array_merge(
	array_values(Tokens::$emptyTokens),
	[
	T_INLINE_HTML,
	T_CLOSE_TAG,
	]
	);
	static $empty;
	if (isset($empty) === false) {
	$empty = Tokens::$emptyTokens;
	$empty[T_INLINE_HTML] = T_INLINE_HTML;
	$empty[T_CLOSE_TAG] = T_CLOSE_TAG;
	}
	return $empty;

It's not as if this compound array will change at any time during the run.

[sirbrillig]

Oh! Good to know. I reduced how often this function is called in #236 but I made #237 to make sure I remember to clean this up.