A character outside the Basic Multilingual Plane is written as two
consecutive \u control words holding the halves of a surrogate pair, as
in \uc1\u-10187 ?\u-9157 ?. Each half was converted on its own, and
since Text cannot hold an unpaired surrogate it became U+FFFD, so 𝐻 came
out as two replacement characters.
Hold a high surrogate back until we know whether the low surrogate
completing it follows, and combine the two. An unpaired surrogate
denotes no character, so it is now dropped rather than replaced; any text
other than the next \u ends the wait, so halves separated by text are not
combined.
Closes#11920.
John MacFarlane
2bddb3274b