Filter to valid surrogate pairs

[guybedford]

This adds filtering to ensure that string encodings always correspond to valid surrogate pairs so that only valid UTF-8 codepoints are represented.

[nicolo-ribaudo]

Is having single-file scripts a design goal? Otherwise you could put the decode function in a separate file to avoid duplicating it.

[guybedford]

Is having single-file scripts a design goal? Otherwise you could put the decode function in a separate file to avoid duplicating it.

The JS lexer.js doesn't have a build so it's likely simpler for now.