Allow token selection around a negative lookaround for a regex rule

[danny0838]

Related code: https://github.com/gorhill/uBlock/blob/2204451514f1a894a7627793a253a89ebbc6d845/src/js/static-filtering-parser.js#L3044-L3048

I suspect that '' instead of '\x01' should be returned for a negative lookaround.

A negative lookaround is a restrictive rule and does not add a possible match. We can expect that a URL matching a rule /\.foo(?!bar\.baz)bar\./ must match /\.foobar\./ and thus must contain foobar, and foobar doesn't need to be excluded during token selection.

[danny0838]

@gorhill Thank you for the quick fixing.

I think positive lookarounds may (maybe "should") be treated differently, as it sometimes works like \b to mark a word separator, and may be used with a back reference to mimic an atomic group like (?=(regex))\1.

Actually I used to think the original version of treating is intentional as it works well for something like abc(?=def). Although it is flawed for something like abc(?=def)(?=ghi), it's not likely to happen in real world regexes.

Maybe we can do something like handling quantifiers to insert optional \x00 or \x01.