feat: transform regular expression literals containing non-ASCII characters

[chirsz-ever]

I choose to transform regular expression literals to new RegExp calls instead of only to replace the non-ASCII characters to escape sequence, because the latter will change the return value of tostring of the created regular expression.

> /π/.toString()
'/π/'
> /\u03C0/.toString()
'/\\u03C0/'
> new RegExp('\u03C0').toString()
'/π/'