Use all char points when generating indexes instead of only the first

[TonyStew]

When generating indexes for a given label we currently only take into account the first codepoint of any given char. This can cause issues. For example the common LGR currently provided by ICANN contains:

    <char cp="0073 0073" ref="118" comment="Sequence added for variant mapping">
      <var cp="00DF" type="blocked" ref="118" comment="IDNA2003 Compatibility" />
      <var cp="03B2" type="blocked" ref="118" />
    </char>

This is meant to mark "ß" as a variant of "ss" but because of the current index generation logic any label with "ss" will return an index containing only "s". IE "sharpness" -> "sharpnes". I don't think this is intended behavior and will cause a lot of collisions.

This PR seems to resolve the issue for me ("sharpness" -> "sharpnes", "teßt" -> "tesst") but more rigorous testing may be needed.