Support `bytes.matches(string)` and `matches(bytes, string)`

[rodaine]

CEL supports string.matches(string) to apply an RE2 regular expression to strings, but RE2 patterns can safely be applied to a non-UTF8 bytes string (the pattern itself must be UTF-8, however). Can we update the spec for matches to include this overload?

Context from protovalidate: bufbuild/protovalidate#268

[TristonianJones]

I'd be okay with adding this functionality, though it'd have to be part of one of the extension functions perhaps. What do you think?