+ *
+ * Note: another choice for the design of this function would be to not
+ * copy the resulting fields at all, but to return indexes and
+ * lengths within the source text.
+ * Advantages would be
+ * o Faster. No Copying.
+ * o Nothing extra needed when field data may contain embedded NUL chars.
+ * o Less memory needed if working on large data.
+ * Disadvantages
+ * o Less consistent with C++ split, which copies into an
+ * array of UnicodeStrings.
+ * o No NUL termination, extracted fields would be less convenient
+ * to use in most cases.
+ * o Possible problems in the future, when support Unicode Normalization
+ * could cause the fields to not correspond exactly to
+ * a range of the source text.