Regression in autocomplete for values with doc comments including unicode text.

[cristianoc]

For example Js.String2.toU should autocomplete to Js.String2.toUpperCase but the generated JSON has problems escaping this:

toUpperCase str converts str to upper case using the locale-insensitive case mappings in the Unicode Character Database. Notice that the conversion can expand the number of letters in the result; for example the German ß capitalizes to two Ses in a row.

toUpperCase "abc" = "ABC";;
toUpperCase {js|Straße|js} = {js|STRASSE|js};;
toLowerCase {js|πς|js} = {js|ΠΣ|js};;