Better Unicode support

[refi64]

It's kind of half-way there at the moment, but it also kind of isn't (try using . with a UTF-32 string; it only matches bytes). My idea:

• If RJ_FUNICODE is passed, then compile the code to work in UTF-32 runes, NOT bytes.

• rejit_match can be amended to convert the input string into runes if it detects the regex had been compiled with RJ_FUNICODE.

• A new function will be added, rejit_unicode_match(regex, rune_str). If the regex had NOT been compiled with RJ_FUNICODE, then this will implicitly convert the rune string to a byte string. Might be a bit risky and bordering too much magic, but...eh.

[refi64]

Actually, screw the implicit conversions. It'll just be a runtime error to call rejit_match with a Unicode regex. Reasoning: it would totally screw anything and everything related to positioning (e.g. groups, the actual return value, etc.).