Fix parsing unicode on v1.0

[elia]

backport of #2073

JavaScript works with UTF-16/UCS-2 and the strings we're passing to
the lexer ar not UTF-8, so the lexer tries to unpack them as 8-bit
chars, messing up the source ranges in which to look for pieces of
code and ultimately shifting all source lookups after having
encountered a unicode string.

E.g.:

  # the string '5' is reported at index 10 when using C*, 6 with U*
  "123️⃣45".unpack('C*').index('5'.unpack('C*').first) # => 10
  "123️⃣45".unpack('U*').index('5'.unpack('U*').first) # => 6