There are apparently various places that want to check for a Unicode scalar
value rather than for a Unicode code point.  Changed those uses of
rtl::isUnicodeCodePoint where that was obvious.  (For changing
svtools/source/svrtf/svparser.cxx see 8e0fb74d
"Revert 'svtools: HTML import: don't put lone surrogates in OUString'".)  Other
uses of rtl::isUnicodeCodePoint might also want to use rtl::isUnicodeScalarValue
instead.

As a side effect, this change also introduces rtl::isSurrogate, which is useful
in a few places as well.

Change-Id: I9245f4f98b83877145a4d392f0ddb7c5d824a535