The helper marco is used in export code to convert UTF-16 internal
string representation to UTF-8 used in XML encoding. I suppose that all
strings here should be already valid UTF-16 (an invalid input should
have been validated at import/input stage). An invalid string at this
stage means a programming error in another part of code that breaks this
precondition, and should not be handled at export stage.
(See also commit 0267a232.)

This effectively changes flags used in conversion from UTF-16 to UTF-8,
so that now RTL_UNICODETOTEXT_FLAGS_INVALID_ERROR is used instead of
RTL_UNICODETOTEXT_FLAGS_INVALID_DEFAULT, so that the result would be
truncated on invalid bytes instead of inserting U+FFFD, and would fail
assertion in debug builds.
Other changed flags don't affect UTF-16-to-UTF-8 conversion.

Change-Id: I12b2cc5378208904c3266924187d6402700ed6f3
Reviewed-on: https://gerrit.libreoffice.org/69801Reviewed-by: Mike Kaganski <mike.kaganski@collabora.com>
Tested-by: Mike Kaganski <mike.kaganski@collabora.com>