#1 Fix CHARACTERS_TOKENs that cross buffer boundry

[Robert G. Jakabosky]

Patch to return PREMATURE_END_OF_BUFFER, Instead of returning a partial CHARACTERS_TOKEN when text is split by the end of a buffer.

Example:
First buffer holds:
<?xml version="1.0" encoding="UTF-8"?>
<doc>
<name>long_name_split

Second buffer holds:
_by_end_of_buffer</name>
</doc>

[Robert G. Jakabosky]

new version of patch.

[Robert G. Jakabosky]

example program to show the problem.

[Robert G. Jakabosky]

Updated the patch to correct it. Also attached an example program to show how CHARACTERS_TOKEN's get fragmented.